Afrikagglers, the African kagglers

Introducción

En este informe se presentan los metodos y resultados de una encuesta realizada a la comunidad de Kaggle sobre sus “usos” en aprendizaje automático y ciencia de datos. Lo anterior, con el objetivo de ofrecer una visión general de los Afrikagglers, como les gusta llamarse, y a su vez resaltar sus experiencias en ciencia de datos y aprendizaje automático. Además, de determinar cómo se compara África con los otros continentes.

Importe de las bibliotecas necesarias

library(tidyverse)
library(grid)
library(gridExtra)
library(ggforce)

Bases de datos

En este informe se usaran las siguientes bases de datos:

• El 2017 Machine Learning and Data Science survey

• El 2018 Machine Learning and Data Science survey

• Un conjunto de datos personalizado con la lista de países presentes en la encuesta de 2018 y cada continente correspondiente.

# STRING PROCESSING
# countries
multipleChoice18$Q3 <- str_replace(multipleChoice18$Q3,"Iran, Islamic Republic of...","Iran")
multipleChoice18$Q3 <- str_replace(multipleChoice18$Q3,"I do not wish to disclose my location","Won't disclose")
multipleChoice18$Q3 <- str_replace(multipleChoice18$Q3,"United Kingdom of Great Britain and Northern Ireland","UK and NI")
multipleChoice18$Q3 <- str_replace(multipleChoice18$Q3,"United States of America","USA")

continents$Country <- str_replace(continents$Country,"Iran, Islamic Republic of...","Iran")
continents$Country <- str_replace(continents$Country,"I do not wish to disclose my location","Won't disclose")
continents$Country <- str_replace(continents$Country,"United Kingdom of Great Britain and Northern Ireland","UK and NI")
continents$Country <- str_replace(continents$Country,"United States of America","USA")

# CONVERT CATEGORICAL DATA TO FACTOR
# age groups
multipleChoice18$Q2 <- factor(multipleChoice18$Q2,
                              levels = c("18-21","22-24","25-29",
                                        "30-34","35-39","40-44",
                                        "45-49","50-54","55-59",
                                        "60-69","70-79","80+"), 
                           labels = c("18-21","22-24","25-29",
                                      "30-34","35-39","40-44",
                                      "45-49","50-54","55-59",
                                      "60-69","70-79","80+"))

# degree
multipleChoice18$Q4 <- factor(multipleChoice18$Q4,
                              levels = c("Doctoral degree","Master’s degree","Bachelor’s degree","Professional degree",
                                         "No formal education past high school",
                                         "Some college/university study without earning a bachelor’s degree",
                                         "I prefer not to answer"), 
                           labels = c("PhD","Master","Bachelor","Professional",
                                         "High school","No degree","Won't disclose"))

# undergraduate major
multipleChoice18$Q5 <- factor(multipleChoice18$Q5, 
                               levels = c("Medical or life sciences (biology, chemistry, medicine, etc.)",
                                          "Computer science (software engineering, etc.)",
                                          "Engineering (non-computer focused)",
                                          "Mathematics or statistics",
                                          "A business discipline (accounting, economics, finance, etc.)",
                                          "Environmental science or geology",
                                          "Social sciences (anthropology, psychology, sociology, etc.)",
                                          "Physics or astronomy",
                                          "Information technology, networking, or system administration",
                                          "I never declared a major",
                                          "Other",
                                          "Humanities (history, literature, philosophy, etc.)") ,
                               labels = c("Medical/life sciences", "Computer science",
                                          "Engineering", "Mathematics/statistics",
                                          "A business discipline", "Physics/astronomy",
                                          "IT/Network/Sys. admin", "No major declared",
                                          "Humanities", "Env. science", "Social sciences", "Other"))

# In what industry is your current employer?
multipleChoice18$Q7 <- factor(multipleChoice18$Q7,
                              levels = c("Retail/Sales", "I am a student", 
                                         "Computers/Technology", "Accounting/Finance",
                                         "Academics/Education", 
                                         "Insurance/Risk Assessment","Other",
                                         "Energy/Mining", "Non-profit/Service",
                                         "Marketing/CRM", "Government/Public Service",
                                         "Manufacturing/Fabrication", 
                                         "Online Service/Internet-based Services",
                                         "Broadcasting/Communications",
                                         "Medical/Pharmaceutical",
                                         "Online Business/Internet-based Sales",
                                         "Military/Security/Defense",
                                         "Shipping/Transportation",
                                         "Hospitality/Entertainment/Sports"),
                              labels = c("Retail / Sales", "Student", 
                                         "Computers / Technology", "Accounting / Finance",
                                         "Academics / Education", 
                                         "Insurance / Risk Assessment","Other",
                                         "Energy / Mining", "Non-profit / Service",
                                         "Marketing / CRM", "Government / Public Service",
                                         "Manufacturing / Fabrication", 
                                         "Online Service / Internet-based Services",
                                         "Broadcasting / Communications",
                                         "Medical / Pharmaceutical",
                                         "Online Business / Internet-based Sales",
                                         "Military / Security/Defense",
                                         "Shipping / Transportation",
                                         "Hospitality / Entertainment/Sports"))

# experience in current role
multipleChoice18$Q8 <- factor(multipleChoice18$Q8, levels = c("0-1","1-2","2-3",
                                                   "3-4","4-5","5-10",
                                                   "10-15","15-20","20-25",
                                                   "25-30","30+"))

# yearly compensation
multipleChoice18$Q9 <- factor(multipleChoice18$Q9, 
                         levels = c("I do not wish to disclose my approximate yearly compensation",
                                   "0-10,000","10-20,000","20-30,000","30-40,000",
                                   "40-50,000","50-60,000","60-70,000","70-80,000",
                                   "80-90,000","90-100,000","100-125,000",
                                   "125-150,000","150-200,000","200-250,000",
                                   "250-300,000","300-400,000", "400-500,000","500,000+"),
                         labels = c("Won't disclose",
                                   "0-10,000","10-20,000","20-30,000","30-40,000",
                                   "40-50,000","50-60,000","60-70,000","70-80,000",
                                   "80-90,000","90-100,000","100-125,000",
                                   "125-150,000","150-200,000","200-250,000",
                                   "250-300,000","300-400,000", "400-500,000","500,000+"))

# time spent coding
multipleChoice18$Q23 <- factor(multipleChoice18$Q23, levels = c("0% of my time",
                                                                "1% to 25% of my time",
                                                                "25% to 49% of my time",
                                                                "50% to 74% of my time",
                                                                "75% to 99% of my time",
                                                                "100% of my time"), 
                               labels = c("0%","1% to 25%","25% to 49%",
                                          "50% to 74%","75% to 99%","100%"))
# coding experience
multipleChoice18$Q24 <- factor(multipleChoice18$Q24, 
                               levels = c("I have never written code and I do not want to learn",
                                          "I have never written code but I want to learn",
                                          "< 1 year","1-2 years","3-5 years","5-10 years",
                                          "10-20 years","20-30 years","30-40 years", "40+ years") ,
                               labels = c("I don't write code and don't want to learn",
                                         "I don't write code but want to learn",
                                        "< 1 year", "1-2 years", "3-5 years", 
                                       "5-10 years", "10-20 years","20-30 years","30-40 years", "40+ years")
)

# For how many years have you used machine learning methods
multipleChoice18$Q25 <- factor(multipleChoice18$Q25,
                               levels = c("I have never studied machine learning and I do not plan to", 
                                          "I have never studied machine learning but plan to learn in the future",
                               "< 1 year", "1-2 years", "2-3 years", "3-4 years", "4-5 years", 
                               "5-10 years", "10-15 years", "20+ years"),
                               labels = c("Never studied, do not plan to", 
                                          "Never studied, plan to learn",
                               "< 1 year", "1-2 years", "2-3 years", "3-4 years", "4-5 years", 
                               "5-10 years", "10-15 years", "20+ years"))

# use of machine learning in industries
multipleChoice18$Q10 <- factor(multipleChoice18$Q10,
                               levels = c("I do not know",
                                          "No (we do not use ML methods)",
                                          "We are exploring ML methods (and may one day put a model into production)",
                                          
                                          "We recently started using ML methods (i.e., models in production for less than 2 years)",
                                          "We have well established ML methods (i.e., models in production for more than 2 years)",
                                          "We use ML methods for generating insights (but do not put working models into production)"),
                               labels = c("I do not know", "No", "Exploring ML methods",
                                          "Recently started", "Well established ML methods", 
                                          "For generating insights"))
# expertise in data science
multipleChoice18$Q40 <- factor(multipleChoice18$Q40, 
                            levels = c("Independent projects are equally important as academic achievements",
                                       "Independent projects are much more important than academic achievements",
                                       "Independent projects are slightly more important than academic achievements",
                                       "Independent projects are slightly less important than academic achievements",
                                       "Independent projects are much less important than academic achievements",
                                       "No opinion; I do not know"),
                            labels = c("Equally important",
                                       "Much more important",
                                       "Slightly more important",
                                       "Less important",
                                       "Much less important",
                                       "No opinion/Don't know"))


# are you a data scientist?
multipleChoice18$Q26 <- factor(multipleChoice18$Q26, 
                               levels = c("Definitely yes", "Probably yes", "Maybe", 
                                          "Probably not", "Definitely not"), 
                               labels = c("Definitely yes", "Probably yes", "Maybe", 
                                          "Probably not", "Definitely not"))

Quienes somos

##País de residencia

En la encuesta hay 57 países representados. Los países con menos de 50 encuestados se clasificaron como “Otros” y algunos encuestados optaron por no proporcionar su información.

De todos los países representados en la encuesta, 6 son africanos. En términos del número total de participantes, representan 681 ó 2.85 % del número total de encuestados. Aunque este valor es bastante bajo, sitúa al continente africano en el quinto lugar. Oceanía es el continente menos representado, con sólo 407 encuestados, mientras que Asia tiene la mayor cantidad.

newMultipleChoice %>%
  group_by(Continent) %>%
  summarise(Count = length(Continent)) %>%
  mutate(highlight_flag = ifelse((Continent == "Africa"), T, F)) %>%
  ggplot(aes(x = reorder(Continent,-Count), y = Count, fill = Continent)) +
  geom_bar(aes(fill = highlight_flag), stat = "identity", color = "grey") +
  geom_text(aes(label =as.character(Count)), 
            position = position_dodge(width = 1), 
            hjust = 0.5, vjust = -0.25, size = 3) +
  scale_fill_brewer(palette = "PuBu") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 12),
        axis.text.y = element_text(size = 12), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Number of respondents", 
       x = "", y = "Count", fill = "",
       caption = "Africa and the world")

En comparación con el año anterior, el número de encuestas africanas muestra un aumento significativo de ´r round(100*(nrow(df18)-nrow(df17))/nrow(df17),2)´%.

p1 <- df %>%
  group_by(Country,Year) %>%
  summarise(Count = length(Country)) %>%
  ggplot(aes(x = Year, y = Count, group = Country)) +
  geom_line(aes(color = Country), size = 0.5) +
  geom_point(aes(color = Country), size = 4) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        plot.subtitle = element_text(hjust = 0.5),
        axis.text = element_text(size = 12), 
        legend.position = "bottom",
        legend.title=element_blank(),
        legend.text = element_text(size = 10)) +
  labs(title = "Number of respondents",
       x = "", y = "Count", fill = "", caption = "")

p2 <- afroCountries %>% 
  group_by(Q1,Q3) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  summarise(Count = length(Q3)) %>%
  ggplot(aes(x = reorder(Q3,-Count), y = Count, fill = Q1)) +
  geom_bar(stat = "identity") +
  scale_fill_brewer(palette = "Paired") + 
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 12), 
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.position = "top",
        legend.text = element_text(size = 10)) +
  labs(title = "Country of residence", x = "", y = "", fill = "",
       caption = "About us")

grid.arrange(p1,p2, ncol = 2)

Además del importante aumento de las encuestados en nigerianas, se han introducido nuevas encuestados procedentes de Marruecos y Túnez.

Distribución de género.

El gráfico del lado derecho de la figura anterior muestra que el número de mujeres es relativamente bajo en comparación con el de hombres. Esto no es sorprendente ya que se aplica a todos los países. Sin embargo, al comparar la proporción de mujeres y hombres en todos los países, me sorprendió ver:

• Túnez tiene una mayor proporción de mujeres que de hombres, y

• Cinco de los seis países africanos incluidos en esta encuesta tienen la mayor proporción de mujeres y hombres entre los 15 países.

multipleChoice18 %>%
  group_by(Q1,Q3) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q3)) %>%
  summarise(Count = n()) %>%
  spread(Q1,Count) %>%
  mutate(ratio = Female/Male) %>%
  mutate(highlight_flag = ifelse((Q3 == "Egypt" | Q3 == "Kenya" | Q3 == "Morocco" |
                                   Q3 == "Nigeria" | Q3 == "Tunisia" | Q3 == "South Africa"), T, F)) %>%
  ggplot(aes(x = reorder(Q3,-ratio), y = ratio, fill = ratio)) +
  geom_bar(aes(fill = highlight_flag), stat = "identity") +
  scale_fill_brewer(palette = "Paired") +
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 11),
        axis.text.x = element_text(size = 9.5, angle = -90,
                                   hjust = 0 , vjust = 0.5)) +  
  labs(title = "Female to Male ratio", 
       x = "", y = "Ratio", fill = "",
       caption = "About us")

Nigeria tiene la proporción más baja de mujeres y hombres de África, pero tiene el mayor número de encuestas. En comparación con otros continentes, África ocupa el segundo lugar detrás de América del Norte.

newMultipleChoice %>%
  group_by(Continent,Q1) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  summarise(Count = n()) %>%
  spread(Q1,Count) %>%
  mutate(ratio = Female/Male) %>%
  mutate(highlight_flag = ifelse((Continent == "Africa"), T, F)) %>%
  ggplot(aes(x = reorder(Continent,-ratio), y = ratio, fill = ratio)) +
  geom_bar(aes(fill = highlight_flag), stat = "identity", color = "grey") +
  scale_fill_brewer(palette = "PuBu") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 12),
        axis.text.x = element_text(size = 12)) +  
  labs(title = "Female to Male ratio", 
       x = "", y = "Ratio", fill = "",
       caption = "Africa and the world")

Sin embargo, esto es bastante prometedor.

Distribución por edad

Más de la mitad de los encuestados tienen entre 22 y 29 años, y ningún encuestado tiene más de 69 años. La mayoría de los hombres tienen entre 22 y 24 años, lo que difiere de la distribución de edad promedio, que tiene la mayoría de personas entre 25 y 29 años. 29. La mayoría de las mujeres tienen entre 25 y 29 años, similar a la mayoría de las mujeres en todos los países.

Algunos grupos de edad no incluyen a las mujeres.

temp <- afroCountries %>%
  group_by(Q1,Q2) %>%
  summarise(Count = n())

ggplot(data = temp, 
       aes(x = Q2, fill = Q1)) +
  geom_bar(data = filter(temp, Q1 == "Male"), aes(y = Count), stat = "identity") +
  geom_bar(data = filter(temp, Q1 == "Female"), aes(y = -1*Count), stat = "identity") +
  scale_y_continuous(breaks = seq(-50,150,50), 
                     labels = as.character(c(seq(50,0,-50), seq(50,150,50)))) + 
  scale_fill_brewer(palette = "Paired") +
  coord_flip() +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text =   element_text(size = 12),
        legend.position = "top",
        legend.text = element_text(size = 11)) + 
  labs(title = "Age distribution in Africa", 
       x = "Age group (years)", y = "Count", fill = "", 
       caption = "")

Aproximadamente el 75% de los Afrikagglers tienen menos de 29 años.

Antecedentes educativos

Graduado

África es el único continente donde las licencias se obtienen con mayor frecuencia. En todos los demás continentes los másteres son los más comunes, sobre todo en Europa, donde el número de doctores también es el mayor.

p1 <- newMultipleChoice %>%
  group_by(Continent,Q4) %>%
  filter(!is.na(Q4)) %>%
  summarise(Count = length(Continent)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = Q4, y = pct, group = Continent)) +
  geom_line(aes(color = Continent), size = 0.5) +
  geom_point(aes(color = Continent), size = 2) +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 5)) +
  scale_y_discrete(labels = function(x) str_wrap(x, width = 30))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 12),
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) +
  labs(title = "Educational background", 
       x = "", y = "%", fill = "",
       caption = "")

p2 <- afroCountries %>%
  group_by(Q1,Q4) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q4)) %>%
  summarise(Count = length(Q4)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = "", y = pct, fill = Q4)) +
  geom_col(width = 1) + 
  scale_fill_brewer(palette = "Set3") + 
  facet_grid(Q1~.) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 12), 
        legend.text = element_text(size = 11)) +
  labs(title = "Degree", 
       x = "", y = "", fill = "Degree",
       caption = "About us")

grid.arrange(p1,p2,ncol = 2)

Sin embargo, en África las mujeres tienen más probabilidades que los hombres de obtener títulos de maestría y doctorado. Además, sólo un pequeño porcentaje de mujeres no tiene títulos, a diferencia de los hombres.

Un alcalde sin título

La carrera universitaria más popular para los Kagglers es la informática. Otras especialidades incluyen matemáticas/estadística e ingeniería.

newMultipleChoice %>%
  group_by(Continent,Q5) %>%
  filter(!is.na(Q5)) %>%
  summarise(Count = length(Q5)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot() + 
  geom_point(mapping = aes(x = Continent, y = reorder(Q5,pct), 
                           size = pct, color = Q5)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 5)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 11), 
        axis.text.x = element_text(size = 12),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Undergraduate major", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

En África, las mujeres tienen más probabilidades que los hombres de haber recibido formación en tecnología de la información. Esto también se aplica a Matemáticas/Estadística. Entonces, ¡buen trabajo, señoras!

Además, una mayor proporción de hombres no ha declarado ninguna especialidad. Esto podría deberse a cierto “no tienen título”.

afroCountries %>%
  group_by(Q1,Q5) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q5)) %>%
  summarise(Count = length(Q5)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = reorder(Q5,-pct), y = pct, group = Q1)) +
  geom_point(aes(color = Q1), size = 2) + geom_line(aes(color = Q1), size = 1) +
  scale_fill_brewer(palette = "Set3") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 11), 
        axis.text.x = element_text(size = 12, angle = -90,hjust = 0,vjust = 0.5),
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) +
  labs(title = "Undergraduate major", 
       x = "", y = "%", fill = "",
       caption = "About us")

Experiencia profesional

Rol actual e industria

África y Asia tienen una mayor proporción de estudiantes. En Europa, América del Norte y América del Sur, la mayoría de las encuestas se centran en la ciencia de datos.

newMultipleChoice %>%
  group_by(Continent,Q6) %>%
  filter(!is.na(Q6)) %>%
  summarise(Count = length(Continent)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot() + 
  geom_point(mapping = aes(x = Continent, y = reorder(Q6,pct), 
                           size = 5*pct, color = Q6)) +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 8)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 12),
        axis.text.y = element_text(size = 10),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Current role", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

Aproximadamente el 25% de los encuestados son estudiantes. A continuación, los trabajos más comunes son los científicos de datos, analistas de datos e ingenieros de software.

afroCountries %>% 
  group_by(Q1,Q6) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q6)) %>%
  summarise(Count = length(Q6)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = reorder(Q6,-pct), y = pct, group = Q1)) +
  geom_point(aes(color = Q1), size = 2) + geom_line(aes(color = Q1), size = 1) +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 12), 
        axis.text.x = element_text(size = 11, angle = -90, 
                                   hjust = 0, vjust = 0.5),
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) +
  labs(title = "Current role", x = "", y = "%", fill = "",
       caption = "About us")

Esta cifra muestra que el porcentaje de estudiantes varones es mayor que el porcentaje de estudiantes mujeres. También podemos ver que las mujeres experimentan menos desempleo.

La mayoría de los periodistas de Data son de Túnez y Nigeria. La mayoría de los científicos de datos residen en Sudáfrica.

afroCountries %>%
  group_by(Q6,Q3) %>% 
  filter(!is.na(Q6)) %>%
  summarise(Count = length(Q6)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q6, y = Q3, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text.x = element_text(size = 12, angle = -90, hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 12), 
        legend.text = element_text(size = 11)) + 
  labs(title = "Current role by country", 
       x = "", y = "", fill = "",
       caption = "About us")

También me gustaría investigar por qué ha habido un aumento en las encuestas africanas desde el año pasado hasta este.

Supongamos que todos los procedimientos fueron realizados por estudiantes, ya que las mujeres que respondieron a la encuesta de 2017 fueron exclusivamente estudiantes. Sin embargo, después de comparar un gráfico circular de estudiantes versus no estudiantes, parece que el porcentaje de estudiantes ha disminuido este año para ambos géneros. Ahora ocupa un lugar más bajo que todas las demás ocupaciones juntas.

propStud17 <- afroCountries17 %>%
  group_by(GenderSelect,StudentStatus) %>%
  filter(GenderSelect == "Female" | GenderSelect == "Male") %>%
  filter(!is.na(StudentStatus)) %>%
  summarise(Count = length(StudentStatus)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = "", y = pct, fill = StudentStatus)) + 
  geom_col(width = 1) +
  coord_polar("y", start = pi / 3) +
  scale_fill_brewer(palette = "Paired") +
  facet_wrap(GenderSelect~.) +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        legend.position = "top") +
  labs(title = "Proportion of students", subtitle = "2017", x = "", y = "", fill = "",
       caption = "")


propStud <- propStud18 %>%
  group_by(Q1,Q6) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q6)) %>%
  summarise(Count = length(Q6)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = "", y = pct, fill = Q6)) +
  geom_col(width = 1) +
  coord_polar("y", start = pi / 3) +
  scale_fill_brewer(palette = "Paired") +
  facet_wrap(Q1~.) +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        legend.position = "top") +
  labs(title = "Proportion of students", subtitle = "2018", x = "", y = "", fill = "",
       caption = "About us")

grid.arrange(propStud17,propStud, ncol = 2)

Esto significa que los profesionales están adoptando un estilo de vida basado en la ciencia de datos.

El aumento está impulsado principalmente por científicos de datos, analistas e ingenieros de software.

# 2017
p1 <- afroCountries17 %>%
  group_by(CurrentJobTitleSelect) %>%
  filter(!is.na(CurrentJobTitleSelect)) %>%
  summarise(Count = length(CurrentJobTitleSelect)) %>%
  ggplot(aes(x = reorder(CurrentJobTitleSelect, Count), y = Count, fill = CurrentJobTitleSelect)) +
  geom_col() +
  coord_flip() + 
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.y = element_text(size = 8),
        axis.text.x = element_text(size = 11),
        legend.position = "none") +
  labs(title = "Current role", subtitle = "2017", x = "", y = "Count", fill = "",
       caption = "")

# 2018
p2 <- afroCountries %>% 
  group_by(Q6) %>%
  filter(Q6 != "Student") %>%
  filter(!is.na(Q6)) %>%
  summarise(Count = length(Q6)) %>%
  ggplot(aes(x = reorder(Q6, Count), y = Count, fill = Q6)) + 
  geom_col() +
  coord_flip()+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.y = element_text(size = 8),
        axis.text.x = element_text(size = 11),
        legend.position = "none") +
  labs(title = "Current role", subtitle = "2018", x = "", y = "Count", fill = "",
       caption = "About us") 

grid.arrange(p1,p2, ncol = 2)

Experiencia en el puesto actual

La mayoría de los estudiantes tienen menos de un año de experiencia, lo cual no es inesperado. Excepto cinco puestos (consultor, director general, director de proyectos, investigador científico y proveedor), la mayoría de los encuestados (> 50%) trabajan en cada categoría durante menos de tres años. Los científicos de datos, periodistas y analistas suelen tener menos de un año de experiencia.

Dada la distribución por edades de los encuestados, esto era de esperar. La mayoría de ellos son jóvenes profesionales.

afroCountries %>%
  group_by(Q6,Q8) %>%
  filter(!is.na(Q6)) %>%
  filter(!is.na(Q8)) %>%
  summarise(Count = length(Q8)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q8, y = Q6, fill = Count)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3) +
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11,
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 9), 
        legend.text = element_text(size = 11)) +
  labs(title = "Experience in current role", 
       x = "Years", y = "", fill = "",
       caption = "About us")

Aplicación del aprendizaje automatizado en la industria.

La mayoría de los estudiantes no comprenden el aprendizaje automatizado en la escuela. Si bien la mayoría de las industrias no utilizan métodos de aprendizaje automatizados, otras están comenzando a experimentar con ellos.

afroCountries %>%
  group_by(Q6,Q10) %>%
  filter(!is.na(Q10)) %>%
  summarise(Count = length(Q10)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q10, y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11), 
        axis.text.y = element_text(size = 10),
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Use of ML in industries", 
       x = "", y = "", fill = "",
       caption = "Machine learning usage")

Most people analyze and understand data to influence business or product decisions. The majority of them complete all tasks at some point, some more than others.

• Research assistants and scientists conduct more research than others.

• Data engineers and database administrators create and manage data infrastructure.

• Data vendors and journalists do not do any of these tasks.

afroCountries %>% 
  select(Q6,Q11_Part_1,Q11_Part_2, Q11_Part_3,Q11_Part_4,Q11_Part_5,Q11_Part_6,Q11_Part_7)%>%
  gather(2:8, key = "questions", value = "Function")%>%
  group_by(Q6,Function)%>%
  filter(!is.na(Function))%>%
  summarise(Count = length(Function))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = Function, y = Q6, fill = pct)) + 
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 9),
        axis.text.y = element_text(size = 9), 
        legend.position = "none") + 
  labs(title = "Day to day function", 
       x = "", y = "", fill = "",
       caption = "Machine learning use")

Experiencia con Codificación

¿Cuánto tiempo dedicas a codificar y analizar datos?

En primer lugar, no hay nadie que no sepa escribir código y no quiera aprender, así que sí! La mayoría, especialmente las mujeres, llevan menos de cinco años codificando. La mayoría de ellos son nuevos y tienen entre uno y dos años de experiencia.

p1 <- afroCountries %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  group_by(Q1,Q24) %>%
  filter(!is.na(Q24)) %>%
  summarise(Count = length(Q24)) %>%
  ggplot(aes(x = Q24, y = Count, fill = Q1)) +
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 12), 
        axis.text.x = element_text(size = 10, angle = -90,hjust = 0,vjust = 0.5),
        legend.position = "top",
        legend.text = element_text(size = 11)) + 
  labs(title = "Coding experience", 
       x = "", y = "Count", fill = "")

p2 <- afroCountries %>%
  group_by(Q3,Q24) %>%
  filter(!is.na(Q3)) %>%
  filter(!is.na(Q24)) %>%
  summarise(Count = length(Q24)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = Q24, y = pct, group = Q3)) +
  geom_point(aes(color = Q3), size = 1.5) + geom_line(aes(color = Q3), size = 0.5) +
  scale_fill_brewer(palette = "Set3") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 15)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 11), 
        axis.text.x = element_text(size = 10, angle = -90,hjust = 0,vjust = 0.5),
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) +
  labs(title = "Coding experience by country", 
       x = "", y = "%", fill = "",
       caption = "About us")

grid.arrange(p1,p2,ncol = 2)

Los ingenieros de datos tienen la más amplia experiencia en codificación, que oscila entre 5 y 10 años. La mayoría de los investigadores tienen de 3 a 5 años de experiencia en codificación.

afroCountries %>%
  group_by(Q6,Q24) %>%
  filter(!is.na(Q24)) %>%
  summarise(Count = length(Q24)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q24, y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90,
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 9), 
        legend.text = element_text(size = 11)) +
  labs(title = "Coding experience", x = "", y = "",
       caption = "Coding experience")

Languages for programming

Most commonly used programming language

The programming language most commonly used by survey respondents is heavily influenced by their current job. Although Python has a clear dominance, the majority of surveys use R. Database engineers prefer SQL. It makes sense; each person chooses what is most suited to their work.

afroCountries %>%
  group_by(Q6,Q17) %>%
  filter(!is.na(Q17)) %>%
  summarise(Count = length(Q17)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = reorder(Q17,-pct), y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3) + 
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = 45, hjust = 1), 
        axis.text.y = element_text(size = 9), 
        legend.text = element_text(size = 11)) +
  labs(title = "Most used programming language", x = "", y = "",
       caption = "Coding experience")

Python suele recomendarse a cualquiera que quiera convertirse en científico de datos. ​

afroCountries %>%
  group_by(Q6,Q18) %>%
  filter(!is.na(Q18)) %>%
  summarise(Count = length(Q18)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = reorder(Q18,-pct), y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3) + 
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11), 
        axis.text.y = element_text(size = 9), 
        legend.text = element_text(size = 11)) +
  labs(title = "Recommended programming language", x = "", y = "",
       caption = "Coding experience")

Incluso aquellos que utilizan habitualmente otro lenguaje para analizar datos recomiendan aprender Python primero. Pero aún así, algunos usuarios fieles de R recomiendan R.

afroCountries %>%
  group_by(Q17,Q18) %>%
  filter(!is.na(Q17)) %>%
  filter(!is.na(Q18)) %>%
  summarise(Count = length(Q17)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = reorder(Q17,pct), y = Q18, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 20))+
  geom_text(aes(label = as.character(Count)), color = "white", size = 3) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 11),
        axis.text.x = element_text(angle = 35, hjust = 1),
        legend.text = element_text(size = 11)) + 
  labs(title = "Most used vs. Recommended programming languages", 
       x = "Most used", y = "Recommended",
       caption = "Coding experience")

Tiempo dedicado a codificar activamente

“La práctica hace la perfección”. Los hombres pasan más tiempo codificando que las mujeres. Estudios anteriores han demostrado que los hombres tienen significativamente más experiencia con la codificación que las mujeres, por lo que tiene sentido. Las mujeres todavía están aprendiendo activamente. Esto también significa que cuanto más código puedas escribir, más tiempo dedicarás a hacerlo.

afroCountries %>%
  group_by(Q1,Q23) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q23)) %>%
  summarise(Count = length(Q23)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = Q23, y = pct, group = Q1)) +
  geom_point(aes(color = Q1), size = 2) + geom_line(aes(color = Q1), size = 1) +
  scale_fill_brewer(palette = "Set3") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.y = element_text(size = 11), 
        axis.text.x = element_text(size = 12), 
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) +
  labs(title = "Time spent actively coding", 
       x = "of time", y = "% of people", fill = "",
       caption = "Coding experience")

Los estudiantes, ingenieros de software, analistas de datos y científicos de datos dedicaron la mayor parte de su tiempo a codificar, entre el 50 y el 74%.

afroCountries %>%
  group_by(Q6,Q23) %>%
  filter(!is.na(Q23)) %>%
  summarise(Count = length(Q23)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q23, y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 11), 
        axis.text.y = element_text(size = 9),
        legend.text = element_text(size = 11)) + 
  labs(title = "Time spent coding", 
       x = "of time", y = "",
       caption = "Coding experience")

IDE

El IDE más utilizado es Jupyter/Ipython. El segundo IDE utilizado por los investigadores es MATLAB, mientras que los estudiantes eligen notepad++. Los ingenieros y analistas de datos eligen RStudio.

afroCountries %>% 
  select(Q6,30:45)%>%
  gather(2:16, key = "questions", value = "IDEs")%>%
  group_by(Q6,IDEs)%>%
  filter(!is.na(IDEs))%>%
  summarise(Count = length(IDEs))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(IDEs,-pct), y = Q6, fill = Count)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90,
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 9), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "IDEs", 
       x = "", y = "", fill = "",
       caption = "Coding experience")

Hosted notebook

La mayoría de los encuestados no utilizan un Hosted notebook. Los kernels de Kaggle son una opción para quienes hacen esto.

afroCountries %>% 
  select(Q6,Q14_Part_1:Q14_Part_11)%>%
  gather(2:12, key = "questions", value = "Hosted_Notebook")%>%
  group_by(Q6,Hosted_Notebook)%>%
  filter(!is.na(Hosted_Notebook))%>%
  summarise(Count = length(Hosted_Notebook))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(Hosted_Notebook,-pct), y = Q6, fill = Count)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 9),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Hosted notebook used at school or work", 
       subtitle = "(past 5 years)",
       x = "", y = "", fill = "",
       caption = "") 

Herramienta principal para el análisis de datos.

La mayoría de los encuestados, en particular estudiantes y científicos de datos, prefieren analizar datos en entornos de desarrollo local.

afroCountries %>%
  group_by(Q6,Q12_MULTIPLE_CHOICE) %>%
  filter(!is.na(Q12_MULTIPLE_CHOICE)) %>%
  summarise(Count = length(Q12_MULTIPLE_CHOICE)) %>%
  ggplot(aes(x = reorder(Q12_MULTIPLE_CHOICE,-Count), y = Q6, fill = Count)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 11), 
        axis.text.y = element_text(size = 9),
        legend.text = element_text(size = 11)) + 
  labs(title = "Primary tools for data analysis", 
       x = "", y = "",
       caption = "Coding experience")

¿Es usted un científico de datos?

Dado que se dedica todo el tiempo a codificar y analizar datos, la mayoría de las encuestas africanas tienden a considerarse científicas. Muchos encuestados son científicos de datos, por lo que esto era de esperar. Pocas mujeres no creen en los datos científicos.

afroCountries %>%
  filter(Q1 == "Female"| Q1 == "Male") %>%
  group_by(Q1,Q26) %>%
  filter(!is.na(Q26)) %>%
  summarise(Count = length(Q26)) %>%
  ggplot(aes(x = Q26, y = Count, fill = Q1))+
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 8)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 9),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.text = element_text(size = 11)) + 
  labs(title = "Think of themself as a data scientist", 
       x = "", y = "Count", fill = "",
       caption = "Personal views")

En comparación con otros continentes, los Afrikagglers tienen más confianza en su identidad de científicos de datos: el 59,89% respondió “Sí”.

newMultipleChoice %>%
  group_by(Continent,Q26) %>%
  filter(!is.na(Q26)) %>%
  summarise(Count = length(Q26)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q26, y = pct, group = Continent)) +
  geom_line(aes(color = Continent), size = 0.5) +
  geom_point(aes(color = Continent), size = 2) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 12),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) + 
  labs(title = "Think of themself as a data scientist", 
       x = "", y = "%", fill = "",
       caption = "Africa and the world")

Uso de métodos de aprendizaje automatizados en el trabajo o en la escuela

Recientemente, los Afrikagglers comenzaron a utilizar métodos de aprendizaje automatizados. La mayoría lo ha utilizado durante menos de un año.

afroCountries %>%
  group_by(Q6,Q25) %>%
  filter(!is.na(Q25)) %>%
  summarise(Count = length(Q25)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q25, y = Q6, fill = Count)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0, vjust = 0.5), 
        axis.text.y = element_text(size = 9),
        legend.text = element_text(size = 11)) + 
  labs(title = "Usage of machine learning at work/school", 
       x = "", y = "",
       caption = "Coding experience")

Productos para el aprendizaje automatizado

Lo mismo ocurre con los productos de aprendizaje automatizado. La mayoría de los encuestados nunca han utilizado una herramienta de aprendizaje automatizado. Sólo unos pocos lo han utilizado, mientras que la mayoría utiliza SAS.

afroCountries %>% 
  select(Q6,152:194)%>%
  gather(2:44, key = "questions", value = "ML_Products")%>%
  group_by(Q6,ML_Products)%>%
  filter(!is.na(ML_Products))%>%
  summarise(Count = length(ML_Products))%>%
  mutate(pct =  prop.table(Count)*100)%>% 
  top_n(5,pct) %>%
  ggplot() + 
  geom_point(mapping = aes(x = reorder(ML_Products,-Count), y = Q6, 
                           size = pct, color = ML_Products)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 8, angle = 45, hjust = 1), 
        axis.text.y = element_text(size = 8), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Machine learning products (past 5 years)", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

Unfortunately, Africa is not alone in this situation. Approximately 25% of respondents have not used any automated learning products.

newMultipleChoice %>% 
  select(Continent,152:194)%>%
  gather(2:44, key = "questions", value = "ML_Products")%>%
  group_by(Continent,ML_Products)%>%
  filter(!is.na(ML_Products))%>%
  summarise(Count = length(ML_Products))%>%
  mutate(pct =  prop.table(Count)*100)%>% 
  top_n(5,Count) %>%
  ggplot() + 
  geom_point(mapping = aes(x = Continent, y = reorder(ML_Products,pct), 
                           size = pct, color = ML_Products)) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 12),
        axis.text.y = element_text(size = 8), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Machine learning products (past 5 years)", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

El uso de productos de aprendizaje automatizado varía según el continente.

• Los productos más utilizados en África son Azure Machine Learning Studio, SAS y Cloudera.

• A los encuestados asiáticos les gusta la API de voz a texto de Google Cloud, la API de Google Cloud Vision y la API de lenguaje natural.

• Los usuarios de Europa utilizan Cloudera, Azure Machine Learning Studio y RapidMiner.

El método de aprendizaje automatizado más utilizado

Scikit-Learn, TensorFlow y Keras son marcos de aprendizaje automatizado populares que se alinean con las tendencias globales.

afroCountries %>% 
  select(Q6,Q19_Part_1:Q19_Part_19)%>%
  gather(2:19, key = "questions", value = "ML_Framework")%>%
  group_by(Q6,ML_Framework)%>%
  filter(!is.na(ML_Framework))%>%
  summarise(Count = length(ML_Framework))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(ML_Framework,-Count), y = Q6, fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 10, angle = 45, hjust = 1),
        axis.text.y = element_text(size = 10), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Machine learning framework (past 5 years)", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

¿Crees que el aprendizaje automatizado es una “caja negra”?

Hay una mayor proporción de mujeres que tienen confianza en sí mismas, pero también una mayor proporción de mujeres que ven el aprendizaje automatizado como cajas negras.

afroCountries %>%
  group_by(Q1,Q48) %>% 
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q48)) %>%
  filter(!is.na(Q1)) %>%
  ggplot(aes(x = Q1, fill = Q48)) +
  geom_bar(position = "fill") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  scale_fill_brewer(palette = "Set3") +
  coord_flip() +
  labs(title = "Do you consider ML as 'black boxes'?", 
       x = "", y = "", fill = "", caption = "Personal views") +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        legend.position = "bottom", 
        axis.text = element_text(size = 12),
        legend.text = element_text(size = 10)) +
  guides(fill = guide_legend(ncol = 1)) 

Métricas utilizadas para determinar el éxito

Las métricas utilizadas por las industrias africanas priorizan la precisión (con más o menos el mismo porcentaje), seguida de los ingresos y/o los objetivos comerciales. Marruecos tiene el mayor porcentaje de personas que no están involucradas en la organización y crean modelos de ML para su empresa.

Sudáfrica y Kenia están más orientadas a los ingresos que otras ciudades. Esto podría explicar las tasas de compensación más altas en estos países (consulte los párrafos siguientes). La injusticia sólo se considera a la ligera.

afroCountries %>% 
  select(Q42_Part_1:Q42_Part_5,Q3) %>%
  gather(1:5, key = "questions", value = "metrics")%>%
  group_by(metrics,Q3) %>%
  filter(!is.na(metrics)) %>%
  filter(!is.na(Q3)) %>%
  ggplot(aes(x = Q3, fill = metrics)) +
  geom_bar(position = "fill") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 15)) +
  scale_fill_brewer(palette = "Set3") +
  labs(title = "Metrics used to measure model success", 
       x = "", y = "", fill = "", caption = "Machine learning usage") +
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 12),
        axis.text.y = element_text(size = 11), 
        legend.text = element_text(size = 11), 
        legend.position = "bottom") +
  guides(fill = guide_legend(ncol = 1))

Este es también el caso de todos los países/continentes.

newMultipleChoice %>% 
  select(Continent,Q42_Part_1:Q42_Part_5)%>%
  gather(2:6, key = "questions", value = "Metrics")%>%
  group_by(Continent,Metrics)%>%
  filter(!is.na(Metrics))%>%
  summarise(Count = length(Metrics))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = Continent, y = reorder(Metrics,pct), fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), 
            hjust = 0.5,vjust = 0.5, size = 4, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  scale_y_discrete(labels = function(x) str_wrap(x, width = 20))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 12),
        axis.text.y = element_text(size = 12), 
        legend.position = "none",
        legend.text = element_text(size = 11)) + 
  labs(title = "Metrics used by organizations", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

¿Qué pasa con los datos?

Tipo de datos más utilizados

La mayoría de la gente utiliza datos numéricos, mientras que los científicos de datos trabajan con datos más tabulares.

afroCountries %>% 
  select(Q6,Q32) %>%
  group_by(Q6,Q32) %>%
  filter(!is.na(Q32)) %>%
  summarise(Count = length(Q32)) %>%
  mutate(pct = round(prop.table(Count)*100,2)) %>%
  ggplot(aes(x = reorder(Q32,-Count), y = Q6, fill = pct)) +
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        legend.position = "none", 
        axis.text.y = element_text(size = 11),
        axis.text.x = element_text(size = 12, angle = -90, hjust = 0, vjust = 0.5),
        legend.text = element_text(size = 11)) +
  labs(title = "Most used data types", 
       x = "Type of data", y = "", fill = "", 
       caption = "Data")

¿Dónde puedo encontrar datos disponibles públicamente?

Las fuentes de datos más utilizadas son las plataformas de agregación de datos, seguidas de las búsquedas en Google y Github.

afroCountries %>% 
  select(Q1,266:276) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  gather(2:12, key = "questions", value = "DataSource")%>%
  group_by(Q1,DataSource) %>%
  filter(!is.na(DataSource)) %>%
  summarise(Count = length(DataSource))%>%
  ggplot(aes(x = reorder(DataSource,-Count), y = Count, fill = Q1)) + 
  geom_col() +
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 20)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text = element_text(size = 12),
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.position = "top",
        legend.text = element_text(size = 11)) +
  labs(title = "Sources used to get public datasets", 
       x = "", y = "Count", fill = "",
       caption = "Data")

Otras herramientas y servicios.

Servicios de computación en la nube en el lugar de trabajo/escuela

Muchos encuestados nunca antes habían utilizado los servicios de computación en la nube. Quienes lo hacen utilizan principalmente los servicios en línea de Amazon, Google Cloud Platform o Microsoft Azure.

afroCountries %>% 
  select(Q6,Q15_Part_1:Q15_Part_7)%>%
  gather(2:8, key = "questions", value = "Cloud_services")%>%
  group_by(Q6,Cloud_services)%>%
  filter(!is.na(Cloud_services))%>%
  summarise(Count = length(Cloud_services))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(Cloud_services,-Count), y = Q6, fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 9),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Cloud computing services at work/school", 
       x = "", y = "", fill = "",
       caption = "") 

Productos informáticos en la nube

En general, una gran parte de los encuestados no utiliza productos de computación en la nube, en particular los estudiantes y analistas de datos. Google Compute Engine es la solución de computación en la nube más popular entre quienes la utilizan.

afroCountries %>% 
  select(Q6,Q27_Part_1:Q27_Part_20) %>%
  gather(2:21, key = "questions", value = "cloud")%>%
  group_by(Q6,cloud)%>%
  filter(!is.na(cloud))%>%
  filter(!is.na(Q6))%>%
  summarise(Count = length(cloud))%>%
  mutate(pct = prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(cloud,-Count), y = Q6, fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15))+
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.5, size = 2.5, color = "white") +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text = element_text(size = 10),
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Cloud computing products (past 5 years)", 
       x = "", y = "", fill = "",
       caption = "Cloud computing products")

Las herramientas de visualización más utilizadas

El amor por Matplotlib es fuerte. ggplot2 y Seaborn continúan siguiéndolo, aunque desde la distancia.

afroCountries %>% 
  select(Q1,Q22)%>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  group_by(Q1,Q22)%>%
  filter(!is.na(Q22))%>%
  summarise(Count = length(Q22))%>%
  ggplot(aes(x = reorder(Q22,-Count), y = Count, fill = Q1)) + 
  geom_col() +
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text = element_text(size = 12),
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.text = element_text(size = 11),
        legend.position = "top") +
  labs(title = "Most used vizualisation libraries", 
       x = "", y = "Count", fill = "",
       caption = "Other tools")

La elección de las bibliotecas de visualización depende, por supuesto, del lenguaje de programación utilizado. Los usuarios de R, SQL, SAS/STATA y JavaScript utilizan ggplot2. Los usuarios de Python y Java utilizan Matplotlib.

afroCountries %>% 
  select(Q17,Q22)%>%
  group_by(Q17,Q22)%>%
  filter(!is.na(Q17)) %>%
  filter(!is.na(Q22))%>%
  summarise(Count = length(Q22))%>%
  mutate(pct =  prop.table(Count)*100)%>%
  ggplot(aes(x = reorder(Q22,-Count), y = Q17, fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), 
            hjust = 0.5,vjust = 0.25, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x,width = 10)) +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text.y = element_text(size = 9),
        axis.text.x = element_text(angle = -90, hjust = 0, vjust = 0.5),
        legend.text = element_text(size = 11),
        legend.position = "none") +
  labs(title = "Most used vizualisation library", 
       x = "", y = "Most used programming language", fill = "",
       caption = "Vizualisation libraries")

Base de datos relacional

En los últimos cinco años, los productos de bases de datos relacionales más populares son MySQL, Microsoft SQL Server y PostgreSQL, y ninguna encuesta utiliza Ingres.

afroCountries %>% 
  select(Q1,196:223) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  gather(2:29, key = "questions", value = "RDB_Products") %>%
  group_by(Q1,RDB_Products) %>%
  filter(!is.na(RDB_Products))%>%
  filter(!is.na(Q1))%>%
  summarise(Count = length(RDB_Products))%>%
  ggplot(aes(x = reorder(RDB_Products,-Count), y = Count, fill = Q1)) +
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +  
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text.x = element_text(size = 8, 
                                   angle = -90, hjust = 0, vjust = 0.5),
        legend.position = "top",
        legend.text = element_text(size = 11)) +
  labs(title = "Relational database products (past 5 years)", 
       x = "", y = "Count", fill = "",
       caption = "Relational database")

Productos para análisis/datos de gran tamaño

Muchas encuestas africanas no utilizan ninguna herramienta analítica o de big data. Google BigQuery es el más popular entre quienes utilizan big data y herramientas analíticas, y hay unos cuantos.

afroCountries %>% 
  select(Q1,225:249) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  gather(2:26, key = "questions", value = "BigData_Products") %>%
  group_by(Q1,BigData_Products) %>%
  filter(!is.na(BigData_Products)) %>%
  filter(!is.na(Q1)) %>%
  summarise(Count = length(BigData_Products))%>%
  ggplot(aes(x = reorder(BigData_Products,-Count), y = Count, fill = Q1)) + 
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text.x = element_text(size = 9, angle = -90, hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 12),
        legend.position = "top",
        legend.text = element_text(size = 11)) +
  labs(title = "Big data and analytics tools (past 5 years)", 
       x = "", y = "Count", fill = "",
       caption = "Big data and analytics tools")

Las herramientas analíticas y los macrodatos no eran tan populares en los continentes africano y asiático. En África, casi la mitad de los encuestados nunca los utilizaron.

newMultipleChoice %>% 
  select(Continent,225:249) %>%
  gather(2:26, key = "questions", value = "BigData_Products") %>%
  group_by(Continent,BigData_Products) %>%
  filter(!is.na(BigData_Products)) %>%
  summarise(Count = length(BigData_Products))%>%
  mutate(pct = prop.table(Count)*100)%>%
  ggplot(aes(x = Continent, y = reorder(BigData_Products,pct), fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), 
            hjust = 0.5,vjust = 0.5, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text = element_text(size = 11),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Big data and analytics tools (past 5 years)", 
       x = "", y = "", fill = "",
       caption = "Africa and the world")

Educación avanzada

Porcentaje actual de formación en aprendizaje automatizado y ciencia de datos.

Para mantenerse a la vanguardia, los Afrikagglers utilizan otros medios para conocer las nuevas tendencias en el aprendizaje automatizado y la ciencia de datos.

El porcentaje promedio de aprendizaje automatizado y ciencia de datos es ligeramente mayor para los hombres encuestados en Autodidacta, Cursos Online y Trabajo. Sin embargo, la mediana es mayor entre las mujeres africanas que asisten a la universidad. En África, las mujeres tienen un mayor porcentaje de educación superior que los hombres. Además, un mayor porcentaje de mujeres estudia tecnología de la información y matemáticas/estadística, lo que puede explicar esto. Dicho esto, sólo unos pocos aprenden a través de las competiciones de Kaggle.

p1 <- multipleChoice18 %>%
  select(Q1,Q35_Part_1) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "",y = Q35_Part_1, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "Self-taught", x = "", y = "%")
  
p2 <- multipleChoice18 %>%
  select(Q1,Q35_Part_2) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "", y = Q35_Part_2, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "Online courses", x = "", y = "%")

p3 <- multipleChoice18 %>%
  select(Q1,Q35_Part_3) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "",y = Q35_Part_3, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "Work", x = "", y = "%")
  
p4 <- multipleChoice18 %>%
  select(Q1,Q35_Part_4) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "", y = Q35_Part_4, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "University", x = "", y = "%")

p5 <- multipleChoice18 %>%
  select(Q1,Q35_Part_5) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "",y = Q35_Part_5, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "Kaggle competitions", x = "", y = "%")
  
p6 <- multipleChoice18 %>%
  select(Q1,Q35_Part_6) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  ggplot(aes(x = "", y = Q35_Part_6, fill = Q1)) + 
  geom_boxplot() +
  scale_fill_brewer(palette = "Paired") +
  theme(plot.title = element_text(size = 13, hjust = 0.5),
        legend.text = element_text(size = 9),
        legend.title = element_blank()) +
  labs(title = "Other", x = "", y = "%")

grid.arrange(p1,p2, p3, p4, p5, p6, ncol = 3)

Plataforma donde comenzó o completó cursos de ciencia de datos

afroCountries %>% 
  select(Q3,Q36_Part_1:Q36_Part_13)%>%
  gather(2:14, key = "questions", value = "OnlinePlatform")%>%
  group_by(Q3,OnlinePlatform)%>%
  filter(!is.na(OnlinePlatform))%>%
  summarise(Count = length(OnlinePlatform))%>%
  mutate(pct =  prop.table(Count)*100)%>% 
  ggplot(aes(x = OnlinePlatform, y = pct, group = Q3)) +
  geom_point(aes(color = Q3), size = 2) + geom_line(aes(color = Q3), size = 0.5) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) + 
  labs(title = "Platform for data science courses", 
       x = "", y = "%", fill = "",
       caption = "")

Coursera es la plataforma de aprendizaje más popular, por delante de DataCamp y Udemy.

Este último parece ser más popular entre los usuarios masculinos (en términos de porcentaje) que entre las mujeres. Este también es el caso de Kaggle Learn (quinto clasificado) y edX (sexto clasificado).

afroCountries %>% 
  select(Q1,Q36_Part_1:Q36_Part_13)%>%
  filter(Q1 == "Female"|Q1 == "Male")%>%
  gather(2:14, key = "questions", value = "OnlinePlatform")%>%
  group_by(Q1,OnlinePlatform)%>%
  filter(!is.na(OnlinePlatform))%>%
  summarise(Count = length(OnlinePlatform))%>%
  ggplot(aes(x = reorder(OnlinePlatform,-Count), y = Count, fill = Q1)) + 
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 20))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.text = element_text(size = 11)) + 
  labs(title = "Online platform used for learning", 
       x = "", y = "Count", fill = "",
       caption = "Online learning")

Plataforma en la que pasaste la mayor parte de tu tiempo

Fuentes de medios favoritas

En general, los foros de Kaggle son los medios de comunicación preferidos para la ciencia de datos, y África tiene la proporción más alta (18,41%).

afroCountries %>% 
  select(Q1,Q38_Part_1:Q38_Part_22)%>%
  filter(Q1 == "Female"|Q1 == "Male")%>%
  gather(2:14, key = "questions", value = "OnlinePlatform")%>%
  group_by(Q1,OnlinePlatform)%>%
  filter(!is.na(OnlinePlatform))%>%
  summarise(Count = length(OnlinePlatform))%>%
  ggplot(aes(x = reorder(OnlinePlatform,-Count), y = Count, fill = Q1)) + 
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 20))+
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.text = element_text(size = 11)) + 
  labs(title = "Online platform used for learning", 
       x = "", y = "Count", fill = "",
       caption = "Online learning")

¿Qué piensas sobre…?

¿Qué piensas sobre los MOOC y los campos de entrenamiento en vivo?

Un número significativo de encuestas favorecen el aprendizaje en línea frente a las instituciones tradicionales, especialmente para los profesionales.

afroCountries %>%
  group_by(Q6,Q39_Part_1) %>%
  filter(!is.na(Q39_Part_1)) %>%
  summarise(Count = length(Q39_Part_1)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = reorder(Q39_Part_1,-Count), y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 10)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 11), 
        axis.text.y = element_text(size = 9),
        legend.text = element_text(size = 11)) + 
  labs(title = "Online learning vs. Traditional institution", 
       x = "", y = "",
       caption = "")

Lo mismo ocurre con el bootcamp presencial.

afroCountries %>%
  group_by(Q6,Q39_Part_2) %>%
  filter(!is.na(Q39_Part_2)) %>%
  summarise(Count = length(Q39_Part_2)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = reorder(Q39_Part_2,-pct), y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 10)) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = as.character(Count)), color = "white", size = 3.5) + 
  theme(legend.position = "none",
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text = element_text(size = 11), 
        axis.text.y = element_text(size = 9),
        legend.text = element_text(size = 11)) + 
  labs(title = "In-person bootcamp vs. Traditional institution", 
       x = "", y = "",
       caption = "")

Experiencia en ciencia de datos

La mayoría de los encuestados cree que los proyectos independientes son más importantes que los logros académicos.

afroCountries %>% 
  select(Q1,Q40) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  group_by(Q1,Q40) %>%
  filter(!is.na(Q40)) %>%
  filter(!is.na(Q1)) %>%
  summarise(Count = length(Q40))%>%
  mutate(pct = prop.table(Count)*100) %>%
  ggplot(aes(x = reorder(Q40,-pct), y = pct, fill = Q1)) + 
  geom_col() + 
  scale_fill_brewer(palette = "Paired") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 10))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        axis.text = element_text(size = 12),
        legend.position = "top",
        legend.text = element_text(size = 11)) +
  labs(title = "Independent projects vs. Academic achievements", 
       x = "", y = "Count", fill = "",
       caption = "Expertise in data science")

45.49% of African respondents believe that independent projects are considerably more important than academic achievements.

newMultipleChoice %>% 
  group_by(Continent, Q40)%>%
  filter(!is.na(Q40))%>%
  summarise(Count = length(Q40))%>%
  mutate(pct = prop.table(Count)*100)%>%
  ggplot(aes(x = Continent, y = reorder(Q40,pct), fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), 
            hjust = 0.5,vjust = 0.25, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  scale_y_discrete(labels = function(x) str_wrap(x, width = 30))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.x = element_text(size = 11),
        axis.text.y = element_text(size = 11),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Expertise in data science",
       subtitle = "Independent projects vs. academic achievements",
       x = "", y = "", fill = "",
       caption = "Africa and the world")

¡Hablemos de dinero!

afroCountries %>% 
  group_by(Q1,Q9)%>%
  filter(Q1 == "Female"|Q1 == "Male")%>%
  filter(!is.na(Q9))%>%
  summarise(Count = length(Q9))%>%
  mutate(pct = prop.table(Count)*100)%>%
  ggplot(aes(x = Q9, y = pct, group = Q1)) +
  geom_point(aes(color = Q1), size = 2) + geom_line(aes(color = Q1), size = 1) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5)) +
  
  theme(plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, vjust = 0.5, hjust = 0),
        axis.text.y = element_text(size = 12), 
        legend.position = "top",
        legend.title = element_blank(),
        legend.text = element_text(size = 11)) + 
  scale_fill_brewer(palette = "Paired") +
  labs(title = "Yearly compensation", 
       x = "$", y = "%", fill = "",
       caption = "About us")

The average income of survey respondents ranged from 0 to 10,000 and 10,000 to 20,000 ‘$’** across all countries. Kenia and South Africa are the most well-paid African countries, with more than $300,000 per surveyed male.

afroCountries %>%
  group_by(Q1,Q9,Q3) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q9)) %>%
  summarise(Count = length(Q9)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q3, y = Q9, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), color = "white", size = 2) + 
  facet_grid(Q1~.) +
  coord_flip() +
  scale_y_discrete(labels = function(x) str_wrap(x, width = 35)) +
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90,  
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 11), 
        legend.text = element_text(size = 11)) + 
  labs(title = "Yearly compensation", 
       x = "", y = "$", fill = "",
       caption = "About us")

Independientemente del título, la mayoría de los encuestados ganan menos de 10.000 dólares al año para ambos sexos. A primera vista, los médicos ganan mucho más que sus homólogas.

afroCountries %>%
  group_by(Q1,Q9,Q4) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q4)) %>%
  filter(!is.na(Q9)) %>%
  summarise(Count = length(Q9)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q4, y = Q9, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), color = "white", size = 2) + 
  facet_grid(Q1~.) +
  coord_flip() +
  scale_y_discrete(labels = function(x) str_wrap(x, width = 35)) +
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90,  
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 11), 
        legend.text = element_text(size = 11)) + 
  labs(title = "Yearly compensation by degree", 
       x = "", y = "$", fill = "",
       caption = "About us")

afroCountries %>%
  group_by(Q1,Q9,Q6) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q6)) %>%
  filter(!is.na(Q9)) %>%
  summarise(Count = length(Q9)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q9, y = Q6, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), color = "white", size = 2) +
  facet_grid(Q1~.) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5)) +
  scale_y_discrete(labels = function(x) str_wrap(x, width = 30)) +
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, 
                                   hjust = 0, vjust = 0.5),
        axis.text.y = element_text(size = 7), 
        legend.text = element_text(size = 11)) +
  labs(title = "Yearly compensation vs. current role", 
       x = "$", y = "", fill = "",
       caption = "About us")

Convertirse en científico de datos, estadístico o ingeniero de datos puede generar mayores ganancias. Entonces, parece que el dinero es donde están los datos.

Según la experiencia actual, las mujeres necesitan de 1 a 2 años de experiencia para ganar entre 90 y 100.000 dólares como científicas de datos, mientras que los hombres pueden ganar hasta 500.000 dólares con la misma experiencia.

Los años de experiencia parecen tener poco impacto en la cantidad de dinero que ganan las personas (especialmente los científicos de datos), pero el género sí.

afroCountries %>%
  group_by(Q1,Q9,Q8) %>%
  filter(Q1 == "Female" | Q1 == "Male") %>%
  filter(!is.na(Q8)) %>%
  filter(!is.na(Q9)) %>%
  summarise(Count = length(Q9)) %>%
  mutate(pct = round(prop.table(Count)*100,2))%>%
  ggplot(aes(x = Q9, y = Q8, fill = pct)) +
  geom_tile(size = 0.5, show.legend = TRUE) +
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), color = "white", size = 2) +
  facet_grid(Q1~.) +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5)) +
  scale_y_discrete(labels = function(x) str_wrap(x, width = 10)) +
  theme(legend.position = "none", 
        plot.title = element_text(size = 15, hjust = 0.5), 
        axis.text.x = element_text(size = 11, angle = -90, hjust = 0),
        axis.text.y = element_text(size = 11), 
        legend.text = element_text(size = 11)) +
  labs(title = "Yearly compensation by gender and experience in current role", 
       x = "$", y = "Years of experience", fill = "",
       caption = "About us")

La mayoría de los encuestados africanos ganan menos de 10.000 dólares. En América del Norte y Oceanía, la mayor proporción de personas gana más de 100.000 dólares. En Europa, la mayoría oscila entre 0 y 60.000$.

newMultipleChoice %>% 
  group_by(Continent, Q9)%>%
  filter(!is.na(Q9))%>%
  summarise(Count = length(Q9))%>%
  mutate(pct = prop.table(Count)*100)%>%
  ggplot(aes(x = Continent, y = Q9, fill = pct)) + 
  geom_tile(stat = "identity") + 
  scale_fill_gradient(low = "salmon1", high = "blue") +
  geom_text(aes(label = sprintf("%.2f%%", pct)), 
            hjust = 0.5,vjust = 0.25, size = 3, color = "white") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 5))+
  scale_y_discrete(labels = function(x) str_wrap(x, width = 30))+
  theme(plot.title = element_text(size = 15, hjust = 0.5),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.x = element_text(size = 11),
        axis.text.y = element_text(size = 11),
        legend.position = "none",
        legend.text = element_text(size = 11)) +
  labs(title = "Yearly compensation",
       x = "", y = "", fill = "",
       caption = "Africa and the world")

Conclusion y resumen

• Se realizan encuestas africanas en Egipto, Kenia, Marruecos, Nigeria, Sudáfrica y Túnez.
• Los países africanos tienen una mayor proporción de mujeres que de hombres.
• La mayoría de los encuestados tienen entre 22 y 29 años.
• La inscripción de estudiantes ha disminuido respecto al año anterior.
• Las mujeres encuestadas tienen un mayor nivel educativo que sus homólogos masculinos.
• Los hombres ganan sustancialmente más que las mujeres, aunque tienen los mismos niveles de experiencia.
• La ciencia de datos se encuentra en sus primeras fases.
• Sudáfrica y Kenia tienen los empleos mejor pagados, mientras que la ciencia de datos es una de las más rentables.
• Los participantes tienen experiencia limitada en codificación pero están ansiosos por aprender más. Los encuestados sudafricanos tienen la mayor experiencia con la codificación. La mayoría de los encuestados nigerianos tenían menos de un año de experiencia.
• Coursera es la plataforma más popular.
• Python (por el contrario), R y SQL son los lenguajes de programación más utilizados.
• La mayoría de las encuestas africanas proporcionan datos científicos.

Es un desafío analizar una muestra tan pequeña, pero en general, la ciencia de datos y el aprendizaje automático todavía están dominados por los jóvenes en África. También tiene un futuro brillante porque cada vez más personas se involucran en él. Espero tener más encuestas el próximo año. En general, los resultados tienden a seguir tendencias amplias, como el uso de Python. Sin embargo, África difiere en otros aspectos, como la distribución por género.