Clusterização de dados para descoberta do contexto da corrrelação entre multi/interdisciplinaridade e produção de tecnologias
Métricas: Viviane Alves. Análise estatística e Código fonte: Viviane Schneider
inicio em 20/02/2020
1 Contexto do Estudo
A análise
Objetivo do estudo: O objetivo do estudo é aplicar técnicas de machine learning para estabelecer o contexto dos programas de pós-graduação que possuem correlação positiva entre Perfil multi/interdisciplinar e produção de tecnologias.
Pergunta de pesquisa: Quais as características do contexto contexto dos programas de pós-graduação que possuem correlação positiva entre Perfil multi/interdisciplinar e produção de tecnologias?
Recorte de análise: Dados dos programas de pós-graduação entre os anos 2013 até 2016.
2 Método e materiais
…..
2.1 Fonte de Dados
- Data from Lattes Platform - Extract by Instituto Stela
- Data from Sucupira Platform
library(dplyr)
library(plotly)
library(ggplot2)
library(tidyr)
library(magrittr)
library(plotrix)
library(rgl)
library(car)
library(lubridate)
library(ggplot2)
library(GGally)
library(corrplot)
library(corrgram)
library(ppcor)
library(plyr)
library(ggpubr)
library(fastcluster)
library(cluster)
.Unianalysis = function (x) {
y <- as.data.frame(table(x))
y <- arrange(y, desc(y$Freq))
return(y)
}#Import Data
library(readr)
IMI <- read_delim("IMI.csv", ";", escape_double = FALSE,
trim_ws = TRUE)
IPT <- read_delim("IPT.csv", ";", escape_double = FALSE,
trim_ws = TRUE)
Estados <- read_delim("Estados.csv", ";", escape_double = FALSE, trim_ws = TRUE)
IMI_IPT <- merge(IMI, IPT, by = "CD_PROGRAMA_IES", all.x = TRUE)
write.csv2(IMI_IPT, file = "IMI_IPT.csv", fileEncoding = "UTF-8")3 Tidyng
3.1 DS_DEPENDENCIA_ADMINISTRATIVA
PUBLICA = 0 PRIVADA = 1
3.2 NM_MODALIDADE_PROGRAMA
ACADEMICO = 0 PROFISSIONAL = 1
3.3 CS_STATUS_JURIDICO
MUNICIPAL = 1 ESTADUAL = 2 FEDERAL = 3 PARTICULAR = 4
3.4 Rename Variables
IMI_IPT_Tidy <- read_delim("IMI_IPT_Tidy.csv",
";", escape_double = FALSE, col_types = cols(ANO_INICIO_PROGRAMA = col_number(),
App_prof = col_number(), CD_CONCEITO_PROGRAMA = col_number(),
CP = col_number(), CS_STATUS_JURIDICO = col_number(),
Colabora_cientifica = col_number(),
Count = col_number(), DS_DEPENDENCIA_ADMINISTRATIVA = col_number(),
Forma_discente = col_number(), Forma_docente = col_number(),
Indice_MI = col_number(), Indice_PT = col_number(),
NM_MODALIDADE_PROGRAMA = col_number(),
Patente_prof = col_number(), Produto_prof = col_number(),
Prof_Perm = col_number(), Soma_Aplicativos = col_number(),
Soma_patentes = col_number(), Soma_produtos = col_number()),
trim_ws = TRUE)## Warning: Missing column names filled in: 'X1' [1]IMI_IPT_T1<- IMI_IPT_Tidy %>%
mutate(new_indice_PT = (Soma_Aplicativos+Soma_produtos+Soma_patentes)/Prof_Perm)
M_State <- merge(IMI_IPT_T1, Estados, by = "CD_PROGRAMA_IES", all.x = TRUE)
write.csv2(IMI_IPT_T1, file = "IMI_IPT_T1.csv", fileEncoding = "UTF-8")
write.csv2(M_State, file = "M_State.csv", fileEncoding = "UTF-8")
cor(IMI_IPT_T1$Indice_MI, IMI_IPT_T1$new_indice_PT, method = "pearson")## [1] 0.038403564 With 3 full metrics of Technologies Production
IMI_IPT_Full_IPT <- IMI_IPT_T1 %>%
filter(Soma_Aplicativos > 0, Soma_patentes >0, Soma_patentes > 0)
dim(IMI_IPT_Full_IPT)## [1] 338 22## X1 CD_PROGRAMA_IES Count CS_STATUS_JURIDICO
## Min. : 10.0 Length:338 Min. : 10.0 Min. :1.000
## 1st Qu.: 883.2 Class :character 1st Qu.: 883.2 1st Qu.:3.000
## Median :1434.5 Mode :character Median :1434.5 Median :3.000
## Mean :1459.2 Mean :1459.2 Mean :2.991
## 3rd Qu.:2114.0 3rd Qu.:2114.0 3rd Qu.:3.000
## Max. :2838.0 Max. :2838.0 Max. :4.000
## DS_DEPENDENCIA_ADMINISTRATIVA ANO_INICIO_PROGRAMA NM_MODALIDADE_PROGRAMA
## Min. :0.0000 Min. :1960 Min. :0.0000
## 1st Qu.:0.0000 1st Qu.:1990 1st Qu.:0.0000
## Median :0.0000 Median :2003 Median :0.0000
## Mean :0.1893 Mean :1998 Mean :0.1479
## 3rd Qu.:0.0000 3rd Qu.:2010 3rd Qu.:0.0000
## Max. :1.0000 Max. :2016 Max. :1.0000
## CD_CONCEITO_PROGRAMA Prof_Perm Soma_patentes Soma_produtos
## Min. :3.000 Min. : 8.00 Min. : 1.000 Min. : 0.000
## 1st Qu.:3.000 1st Qu.:13.00 1st Qu.: 1.000 1st Qu.: 0.000
## Median :4.000 Median :16.00 Median : 3.000 Median : 0.000
## Mean :4.175 Mean :19.93 Mean : 6.234 Mean : 1.799
## 3rd Qu.:5.000 3rd Qu.:23.00 3rd Qu.: 7.000 3rd Qu.: 2.000
## Max. :7.000 Max. :92.00 Max. :91.000 Max. :52.000
## Soma_Aplicativos Forma_docente Forma_discente Colabora_cientifica
## Min. : 1.000 Min. : 2 Min. : 1 Min. : 1
## 1st Qu.: 1.000 1st Qu.:274183006 1st Qu.:133720930 1st Qu.:430488258
## Median : 2.000 Median :449849170 Median :322126437 Median :509435752
## Mean : 4.361 Mean :430858176 Mean :340755780 Mean :453461931
## 3rd Qu.: 5.000 3rd Qu.:611111111 3rd Qu.:555220884 3rd Qu.:560495754
## Max. :33.000 Max. :952380952 Max. :868421053 Max. :958333333
## CP Indice_MI Patente_prof
## Min. : 1 Min. : 524877 Min. : 1
## 1st Qu.:359416041 1st Qu.:375793354 1st Qu.: 32526882
## Median :432500000 Median :467942846 Median : 83333333
## Mean :373090280 Mean :440400754 Mean :172446760
## 3rd Qu.:486395474 3rd Qu.:559820849 3rd Qu.:262585812
## Max. :666666667 Max. :808461211 Max. :933333333
## Produto_prof App_prof Indice_PT new_indice_PT
## Min. : 0 Min. : 1 Min. : 1 Min. :0.05556
## 1st Qu.: 0 1st Qu.: 13313008 1st Qu.: 57045945 1st Qu.:0.25232
## Median : 0 Median : 71428571 Median :128205128 Median :0.47214
## Mean : 47760613 Mean :136148027 Mean :172084562 Mean :0.69475
## 3rd Qu.: 68390804 3rd Qu.:166666667 3rd Qu.:252258356 3rd Qu.:0.88052
## Max. :769230769 Max. :956521739 Max. :743589744 Max. :5.07143
5 Start program after 2002 - Original metrics
## [1] 1723 22## X1 CD_PROGRAMA_IES Count CS_STATUS_JURIDICO
## Min. : 3.0 Length:1723 Min. : 3.0 Min. :1.000
## 1st Qu.: 647.5 Class :character 1st Qu.: 647.5 1st Qu.:3.000
## Median :1390.0 Mode :character Median :1390.0 Median :3.000
## Mean :1422.3 Mean :1422.3 Mean :2.952
## 3rd Qu.:2181.5 3rd Qu.:2181.5 3rd Qu.:3.000
## Max. :2858.0 Max. :2858.0 Max. :4.000
## DS_DEPENDENCIA_ADMINISTRATIVA ANO_INICIO_PROGRAMA NM_MODALIDADE_PROGRAMA
## Min. :0.0000 Min. :2003 Min. :0.0000
## 1st Qu.:0.0000 1st Qu.:2007 1st Qu.:0.0000
## Median :0.0000 Median :2011 Median :0.0000
## Mean :0.1793 Mean :2010 Mean :0.1741
## 3rd Qu.:0.0000 3rd Qu.:2013 3rd Qu.:0.0000
## Max. :1.0000 Max. :2016 Max. :1.0000
## CD_CONCEITO_PROGRAMA Prof_Perm Soma_patentes Soma_produtos
## Min. :3.000 Min. : 2.00 Min. : 0.000 Min. : 0.0000
## 1st Qu.:3.000 1st Qu.:12.00 1st Qu.: 0.000 1st Qu.: 0.0000
## Median :3.000 Median :15.00 Median : 0.000 Median : 0.0000
## Mean :3.439 Mean :15.75 Mean : 1.457 Mean : 0.6019
## 3rd Qu.:4.000 3rd Qu.:18.00 3rd Qu.: 1.000 3rd Qu.: 0.0000
## Max. :6.000 Max. :99.00 Max. :67.000 Max. :52.0000
## Soma_Aplicativos Forma_docente Forma_discente Colabora_cientifica
## Min. : 0.0000 Min. : 1 Min. : 1 Min. : 1
## 1st Qu.: 0.0000 1st Qu.:333333333 1st Qu.:167948718 1st Qu.:433125659
## Median : 0.0000 Median :512820513 Median :447368421 Median :536096257
## Mean : 0.9077 Mean :466274268 Mean :405073136 Mean :464855509
## 3rd Qu.: 1.0000 3rd Qu.:641025641 3rd Qu.:615384615 3rd Qu.:594112850
## Max. :50.0000 Max. :962962963 Max. :978723404 Max. :994318182
## CP Indice_MI Patente_prof
## Min. : 1 Min. : 6125 Min. : 0
## 1st Qu.:366987180 1st Qu.:438288990 1st Qu.: 0
## Median :435964912 Median :521434661 Median : 0
## Mean :382543317 Mean :481936051 Mean : 48311671
## 3rd Qu.:492189542 3rd Qu.:589073952 3rd Qu.: 6
## Max. :833333333 Max. :832421024 Max. :933333333
## Produto_prof App_prof Indice_PT new_indice_PT
## Min. : 0 Min. : 0 Min. : 0 Min. :0.00000
## 1st Qu.: 0 1st Qu.: 0 1st Qu.: 0 1st Qu.:0.00000
## Median : 0 Median : 0 Median : 2 Median :0.03571
## Mean : 22194010 Mean : 36585254 Mean : 51459919 Mean :0.20055
## 3rd Qu.: 0 3rd Qu.: 1 3rd Qu.: 47619048 3rd Qu.:0.18750
## Max. :933333333 Max. :956521739 Max. :743589744 Max. :4.50000
## [1] 0.06924649## [1] 0.06073269## Indice Formação docente
cor(IMI_IPT_M2002$Forma_docente, IMI_IPT_M2002$new_indice_PT, method = "pearson")## [1] 0.09705732## Indice Formação discente
cor(IMI_IPT_M2002$Forma_discente, IMI_IPT_M2002$new_indice_PT, method = "pearson")## [1] 0.004790162## Indice Calaboração cientifica
cor(IMI_IPT_M2002$Colabora_cientifica, IMI_IPT_M2002$new_indice_PT, method = "pearson")## [1] -0.02899132## Indice contexto profissional
cor(IMI_IPT_M2002$CP, IMI_IPT_M2002$new_indice_PT, method = "pearson")## [1] -0.055416036 Correlation by private universities programs
## [1] 496 22## X1 CD_PROGRAMA_IES Count CS_STATUS_JURIDICO
## Min. : 126 Length:496 Min. : 126 Min. :4
## 1st Qu.:1450 Class :character 1st Qu.:1450 1st Qu.:4
## Median :1876 Mode :character Median :1876 Median :4
## Mean :1865 Mean :1865 Mean :4
## 3rd Qu.:2502 3rd Qu.:2502 3rd Qu.:4
## Max. :2857 Max. :2857 Max. :4
## DS_DEPENDENCIA_ADMINISTRATIVA ANO_INICIO_PROGRAMA NM_MODALIDADE_PROGRAMA
## Min. :1 Min. :1963 Min. :0.0000
## 1st Qu.:1 1st Qu.:1999 1st Qu.:0.0000
## Median :1 Median :2006 Median :0.0000
## Mean :1 Mean :2003 Mean :0.2097
## 3rd Qu.:1 3rd Qu.:2011 3rd Qu.:0.0000
## Max. :1 Max. :2016 Max. :1.0000
## CD_CONCEITO_PROGRAMA Prof_Perm Soma_patentes Soma_produtos
## Min. :3.000 Min. : 2.00 Min. : 0.00 Min. : 0.0000
## 1st Qu.:3.000 1st Qu.: 11.00 1st Qu.: 0.00 1st Qu.: 0.0000
## Median :4.000 Median : 13.00 Median : 0.00 Median : 0.0000
## Mean :3.875 Mean : 13.51 Mean : 1.51 Mean : 0.9294
## 3rd Qu.:4.000 3rd Qu.: 15.00 3rd Qu.: 1.00 3rd Qu.: 1.0000
## Max. :7.000 Max. :103.00 Max. :63.00 Max. :52.0000
## Soma_Aplicativos Forma_docente Forma_discente Colabora_cientifica
## Min. : 0.000 Min. : 1 Min. : 1 Min. : 1
## 1st Qu.: 0.000 1st Qu.:365909091 1st Qu.: 63928550 1st Qu.:389921576
## Median : 0.000 Median :538461538 Median :414855072 Median :528257275
## Mean : 1.312 Mean :495290813 Mean :371258017 Mean :443588533
## 3rd Qu.: 1.000 3rd Qu.:692307692 3rd Qu.:581317204 3rd Qu.:583333333
## Max. :24.000 Max. :962962963 Max. :952380952 Max. :992857143
## CP Indice_MI Patente_prof
## Min. : 1 Min. : 445625 Min. : 0
## 1st Qu.:352967172 1st Qu.:438190576 1st Qu.: 0
## Median :441021826 Median :531754331 Median : 0
## Mean :383590377 Mean :494575120 Mean : 50570792
## 3rd Qu.:498999452 3rd Qu.:601939816 3rd Qu.: 1
## Max. :833333333 Max. :832421024 Max. :833333333
## Produto_prof App_prof Indice_PT new_indice_PT
## Min. : 0 Min. : 0 Min. : 0 Min. :0.00000
## 1st Qu.: 0 1st Qu.: 0 1st Qu.: 0 1st Qu.:0.00000
## Median : 0 Median : 0 Median : 11084906 Median :0.07692
## Mean : 30257071 Mean : 46952321 Mean : 71338997 Mean :0.28732
## 3rd Qu.: 1 3rd Qu.: 5 3rd Qu.: 83333333 3rd Qu.:0.27976
## Max. :928571429 Max. :888888889 Max. :743589744 Max. :4.50000## Warning in cor(IMI_IPT_PRIV[, 3:22]): o desvio padrão é zero
## [1] 0.1384989## [1] 0.1135799## Indice Formação docente
cor(IMI_IPT_PRIV$Forma_docente, IMI_IPT_PRIV$new_indice_PT, method = "pearson")## [1] 0.0764407## Indice Formação discente
cor(IMI_IPT_PRIV$Forma_discente, IMI_IPT_PRIV$new_indice_PT, method = "pearson")## [1] 0.00449351## Indice Calaboração cientifica
cor(IMI_IPT_PRIV$Colabora_cientifica, IMI_IPT_PRIV$new_indice_PT, method = "pearson")## [1] -0.07243716## Indice contexto profissional
cor(IMI_IPT_PRIV$CP, IMI_IPT_PRIV$new_indice_PT, method = "pearson")## [1] -0.06698494## Variaveis
M <- IMI_IPT_T1
# mu da amostra de produção de aplicativo
Media_Indice_MI <- mean(M$Indice_MI, na.rm = TRUE)
## mu de diversidade do universo: mu
Media_Prod_tecno <- mean(M$Indice_PT, na.rm = TRUE)
Context_m_acima <-
M %>%
dplyr::filter(Indice_MI > Media_Indice_MI,
Indice_PT > Media_Prod_tecno)
summary(Context_m_acima)## X1 CD_PROGRAMA_IES Count CS_STATUS_JURIDICO
## Min. : 4 Length:449 Min. : 4 Min. :1.000
## 1st Qu.: 728 Class :character 1st Qu.: 728 1st Qu.:3.000
## Median :1452 Mode :character Median :1452 Median :3.000
## Mean :1471 Mean :1471 Mean :3.134
## 3rd Qu.:2174 3rd Qu.:2174 3rd Qu.:4.000
## Max. :2856 Max. :2856 Max. :4.000
## DS_DEPENDENCIA_ADMINISTRATIVA ANO_INICIO_PROGRAMA NM_MODALIDADE_PROGRAMA
## Min. :0.0000 Min. :1963 Min. :0.0000
## 1st Qu.:0.0000 1st Qu.:2001 1st Qu.:0.0000
## Median :0.0000 Median :2008 Median :0.0000
## Mean :0.2918 Mean :2004 Mean :0.2205
## 3rd Qu.:1.0000 3rd Qu.:2012 3rd Qu.:0.0000
## Max. :1.0000 Max. :2016 Max. :1.0000
## CD_CONCEITO_PROGRAMA Prof_Perm Soma_patentes Soma_produtos
## Min. :3.000 Min. : 6.00 Min. : 0.000 Min. : 0.000
## 1st Qu.:3.000 1st Qu.:12.00 1st Qu.: 0.000 1st Qu.: 0.000
## Median :4.000 Median :14.00 Median : 2.000 Median : 0.000
## Mean :3.755 Mean :15.19 Mean : 5.065 Mean : 1.808
## 3rd Qu.:4.000 3rd Qu.:17.00 3rd Qu.: 7.000 3rd Qu.: 2.000
## Max. :7.000 Max. :47.00 Max. :70.000 Max. :52.000
## Soma_Aplicativos Forma_docente Forma_discente Colabora_cientifica
## Min. : 0.000 Min. : 2 Min. : 1 Min. : 1
## 1st Qu.: 0.000 1st Qu.:444444444 1st Qu.:272727273 1st Qu.:454134178
## Median : 1.000 Median :571428571 Median :486486486 Median :531746032
## Mean : 3.096 Mean :524555524 Mean :425949557 Mean :453353999
## 3rd Qu.: 4.000 3rd Qu.:666666667 3rd Qu.:619047619 3rd Qu.:586538056
## Max. :50.000 Max. :952380952 Max. :928571429 Max. :986111111
## CP Indice_MI Patente_prof
## Min. : 1 Min. :449649557 Min. : 0
## 1st Qu.:375757576 1st Qu.:499141484 1st Qu.: 0
## Median :443055556 Median :551957071 Median : 45454545
## Mean :387971735 Mean :559362369 Mean :157810467
## 3rd Qu.:501041667 3rd Qu.:610416667 3rd Qu.:263157895
## Max. :833333333 Max. :827802579 Max. :916666667
## Produto_prof App_prof Indice_PT new_indice_PT
## Min. : 0 Min. : 0 Min. : 55555556 Min. :0.1667
## 1st Qu.: 0 1st Qu.: 0 1st Qu.: 83333333 1st Qu.:0.2500
## Median : 0 Median : 7 Median :148148148 Median :0.4444
## Mean : 69848133 Mean :114844943 Mean :192867856 Mean :0.6816
## 3rd Qu.: 83333333 3rd Qu.:166666667 3rd Qu.:266666667 3rd Qu.:0.8000
## Max. :933333333 Max. :956521739 Max. :743589744 Max. :5.0714
## Indice total original
cor(Context_m_acima$Indice_MI, Context_m_acima$Indice_PT, method = "pearson")## [1] 0.04446282## Indice Total novo
cor(Context_m_acima$Indice_MI, Context_m_acima$new_indice_PT, method = "pearson")## [1] 0.08391618## Indice Formação docente
cor(Context_m_acima$Forma_docente, Context_m_acima$new_indice_PT, method = "pearson")## [1] 0.03300118## Indice Formação discente
cor(Context_m_acima$Forma_discente, Context_m_acima$new_indice_PT, method = "pearson")## [1] -0.06491536## Indice Calaboração cientifica
cor(Context_m_acima$Colabora_cientifica, Context_m_acima$new_indice_PT, method = "pearson")## [1] -0.05040562## Indice contexto profissional
cor(Context_m_acima$CP, Context_m_acima$new_indice_PT, method = "pearson")## [1] -0.11413927 Hierarquical cluster analysis
7.1 Metrics above and start in 2003
8 Non-Hierarquical cluster analysis
8.1 Metrics above and start in 2003
Contex_Full <- M_State %>% dplyr::select(CD_PROGRAMA_IES, Indice_MI, new_indice_PT)
Contex_Full <- na.omit(Contex_Full) Distances <- dist(Contex_Full) dendrogram <- hclust(Distances) plot(dendrogram) rect.hclust(dendrogram, k=6, border=1:3)
plot(hclust(Distances, method=“ward.D”), main=“ward.D”, hang=-1) rect.hclust(hclust(Distances, method=“ward.D”), k=5, border=1:5)
plot(hclust(Distances, method=“centroid”), main=“centroid”, hang=-1) rect.hclust(hclust(Distances, method=“centroid”), k=5, border=1:5)
Kmeans_HW<-kmeans(Contex_Full,100 , iter.max = 5, nstart = 25, algorithm = c(“Hartigan-Wong”), trace=FALSE) HartiganWong<- cor(Kmeans_HW\(centers[,2:2], Kmeans_HW\)centers[,3:3] , method = “pearson”)
plot(Contex_Full, col = Kmeans_HW\(cluster) points(Kmeans_HW\)centers, col = 1:2, pch = 8, cex = 2)
Kmeans_Fo<- kmeans(Contex_Full, 10, iter.max = 5, nstart = 1,
algorithm = c("Forgy"), trace=FALSE)
Forgy <- cor(Kmeans_Fo$centers[,2:2], Kmeans_Fo$centers[,3:3] , method = "pearson")
Kmeans_Ma<- kmeans(Contex_Full, 10, iter.max = 5, nstart = 1,
algorithm = c("MacQueen"), trace=FALSE)
MacQueen <- cor(Kmeans_Ma$centers[,2:2], Kmeans_Ma$centers[,3:3] , method = "pearson")
9 Indices Originais
9.1 Universo por estado
By_State <-
M_State %>%
dplyr::group_by(SG_UF_PROGRAMA) %>%
dplyr::summarise(Patentes_total = sum(Soma_patentes),
Produtos_total = sum(Soma_produtos),
Aplicativos_total = sum(Soma_Aplicativos),
Programas_total = n_distinct(CD_PROGRAMA_IES),
Professores_total = sum(Prof_Perm),
Indice_MI_S = sum(Indice_MI)/Programas_total,
Forma_docente = 1 / sum(Forma_docente),
Forma_discente = 1/ sum(Forma_discente),
Colabora_cientifica = 1/ sum(Colabora_cientifica),
Contex_Profissa = 1/ sum(CP)) %>%
mutate(Indice_MI_S = ((Forma_docente+Forma_discente+Colabora_cientifica+Contex_Profissa)/4)/Programas_total,
Indice_PT_S = (Patentes_total+Produtos_total+ Aplicativos_total)/Programas_total)
cor(By_State$Indice_MI_S, By_State$Indice_PT_S, method = "spearman")## [1] -0.3184247## [1] -0.3226988kMeans1 <- function(dados, k=2){
# função que calcula a distância euclidiana euc.dist <- function(x1, x2) sum((x1 - x2) ^ 2)
# labels rotulo = 1:k rownames(dados)[nrow(dados)]= 1
# random centroids for(i in 1:nrow(dados)){ rownames(dados)[i] <- sample(rotulo,1) }
centroids <- colMeans(dados[rownames(dados) == 1, ])
for(j in 2:k){ centroids <- rbind(centroids, colMeans(dados[rownames(dados) == j, ])) }
rownames(centroids) = 1:k # identifica o centroide de cada grupo
for(i in 1:nrow(dados)){ distancias = NULL for(j in 1:k){ distancias[j] = euc.dist(dados[i,], centroids[j,]) } names(distancias) = 1:k # print(distancias) rownames(dados)[i] = as.numeric(names(distancias[distancias == min(distancias)]))
# recalcula as medias
centroids <- colMeans(dados[rownames(dados) == 1, ])
for(z in 2:k){
centroids <- rbind(centroids, colMeans(dados[rownames(dados) == z, ]))
}} # centroids
return(list(centroides = centroids, grupo1 = dados[rownames(dados) == 1, ], grupo2 = dados[rownames(dados) == 2, ], grupo3 = dados[rownames(dados) == 3, ], clusters = as.numeric(rownames(dados)))) }
dados = as.matrix(Context_m_acima[2:20])
exemplo = kMeans1(dados,k=3) # exemplo\(grupo1 # exemplo\)grupo2 # exemplo\(grupo3 plot(dados,col = exemplo\)clusters+1, main=“K-Means result with 3 clusters”, pch=20, cex=2)
Contex_2002_Above <- IMI_IPT_M2002 %>%
dplyr::filter(Indice_MI > Media_Indice_MI,
Indice_PT > Media_Prod_tecno) %>%
dplyr::select(CD_CONCEITO_PROGRAMA, Indice_MI, new_indice_PT)
Distances <- dist(Contex_2002_Above)
dendrogram <- hclust(Distances)
plot(dendrogram)
rect.hclust(dendrogram, k=6, border=1:3)
plot(hclust(Distances, method="ward.D"), main="ward.D", hang=-1)
rect.hclust(hclust(Distances, method="ward.D"), k=5, border=1:5)
plot(hclust(Distances, method="centroid"), main="centroid", hang=-1)
rect.hclust(hclust(Distances, method="centroid"), k=5, border=1:5)
Kmeans_HW<- kmeans(Contex_2002_Above,10 , iter.max = 5, nstart = 1, algorithm = c(“Hartigan-Wong”), trace=FALSE) HartiganWong<- cor(Kmeans_HW\(centers[,2:2], Kmeans_HW\)centers[,3:3] , method = “pearson”)
plot(Contex_2002_Above, col = Kmeans_HW\(cluster) points(Kmeans_HW\)centers, col = 1:2, pch = 8, cex = 2)
Kmeans_Lo<- kmeans(Contex_2002_Above, 10, iter.max = 5, nstart = 1, algorithm = c(“Lloyd”), trace=FALSE) Lloyd <- cor(Kmeans_Lo\(centers[,2:2], Kmeans_Lo\)centers[,3:3] , method = “pearson”)
Kmeans_Fo<- kmeans(Contex_2002_Above, 10, iter.max = 5, nstart = 1, algorithm = c(“Forgy”), trace=FALSE) Forgy <- cor(Kmeans_Fo\(centers[,2:2], Kmeans_Fo\)centers[,3:3] , method = “pearson”)
Kmeans_Ma<- kmeans(Contex_2002_Above, 10, iter.max = 5, nstart = 1, algorithm = c(“MacQueen”), trace=FALSE) MacQueen <- cor(Kmeans_Ma\(centers[,2:2], Kmeans_Ma\)centers[,3:3] , method = “pearson”)
ResultsCor <- rbind(HartiganWong, Lloyd, Forgy,MacQueen) ResultsCor
Kmeans_Lo<- kmeans(Contex_2002_Above, 10, iter.max = 5, nstart = 1, algorithm = c(“Lloyd”), trace=FALSE) Lloyd <- cor(Kmeans_Lo\(centers[,2:2], Kmeans_Lo\)centers[,3:3] , method = “pearson”)
ResultsCor <- rbind(HartiganWong, Lloyd, Forgy,MacQueen) ResultsCor