Prediccion de venta con modelo lineal
Irving Saul Torres Mendoza
2024-10-31
1 Cargar lo paquetes necesarios
#2 Cargar base de datos
df<-read.csv("Data/Sample - Superstore.csv")
head(df)## Row.ID Order.ID Order.Date Ship.Date Ship.Mode Customer.ID
## 1 1 CA-2016-152156 11/8/2016 11/11/2016 Second Class CG-12520
## 2 2 CA-2016-152156 11/8/2016 11/11/2016 Second Class CG-12520
## 3 3 CA-2016-138688 6/12/2016 6/16/2016 Second Class DV-13045
## 4 4 US-2015-108966 10/11/2015 10/18/2015 Standard Class SO-20335
## 5 5 US-2015-108966 10/11/2015 10/18/2015 Standard Class SO-20335
## 6 6 CA-2014-115812 6/9/2014 6/14/2014 Standard Class BH-11710
## Customer.Name Segment Country City State
## 1 Claire Gute Consumer United States Henderson Kentucky
## 2 Claire Gute Consumer United States Henderson Kentucky
## 3 Darrin Van Huff Corporate United States Los Angeles California
## 4 Sean O'Donnell Consumer United States Fort Lauderdale Florida
## 5 Sean O'Donnell Consumer United States Fort Lauderdale Florida
## 6 Brosina Hoffman Consumer United States Los Angeles California
## Postal.Code Region Product.ID Category Sub.Category
## 1 42420 South FUR-BO-10001798 Furniture Bookcases
## 2 42420 South FUR-CH-10000454 Furniture Chairs
## 3 90036 West OFF-LA-10000240 Office Supplies Labels
## 4 33311 South FUR-TA-10000577 Furniture Tables
## 5 33311 South OFF-ST-10000760 Office Supplies Storage
## 6 90032 West FUR-FU-10001487 Furniture Furnishings
## Product.Name Sales
## 1 Bush Somerset Collection Bookcase 261.9600
## 2 Hon Deluxe Fabric Upholstered Stacking Chairs, Rounded Back 731.9400
## 3 Self-Adhesive Address Labels for Typewriters by Universal 14.6200
## 4 Bretford CR4500 Series Slim Rectangular Table 957.5775
## 5 Eldon Fold 'N Roll Cart System 22.3680
## 6 Eldon Expressions Wood and Plastic Desk Accessories, Cherry Wood 48.8600
## Quantity Discount Profit
## 1 2 0.00 41.9136
## 2 3 0.00 219.5820
## 3 2 0.00 6.8714
## 4 5 0.45 -383.0310
## 5 2 0.20 2.5164
## 6 7 0.00 14.1694#3 Modificaciones de columnas
df_mod <- df %>% #Llamar a la base de datos
filter(State == "Florida") %>% #ESTE ES EL PASO NUEVO
select(Order.Date, Sales) %>% #Selecconamos dos columnas Order.Date y Sales
mutate(Order.Date = mdy(Order.Date)) %>% #Cambiar a tipo fecha
mutate(Fecha = floor_date(Order.Date, unit = "month")) %>% #Regresamos a inicio de mes
group_by(Fecha) %>% #Agrupa por mes
summarise(Ventas = sum(Sales)) %>% #Hacer la suma
mutate(Mes = seq_along(Fecha)) #Crear columna x para la regresion
head(df_mod)## # A tibble: 6 × 3
## Fecha Ventas Mes
## <date> <dbl> <int>
## 1 2014-01-01 25.2 1
## 2 2014-02-01 199. 2
## 3 2014-03-01 25883. 3
## 4 2014-04-01 737. 4
## 5 2014-05-01 1196. 5
## 6 2014-06-01 63.4 64 Visualizacion
df_mod %>%
ggplot(., aes(x=Mes, y=Ventas)) +
geom_point() +
geom_line() +
geom_smooth(method = "lm", formula = "y ~ x")
5 Grafica de caja y bigotes
#Esta grafica me sirve para encontrar valores atipicos
df_mod %>%
ggplot(., aes(x=Ventas)) +
geom_boxplot()
6 Revisar si la variable y tiene una distribucion aprox normal
df_mod %>%
ggplot(., aes(x=Ventas)) +
geom_histogram(bins=30, aes(y=..density..)) +
geom_density()## Warning: The dot-dot notation (`..density..`) was deprecated in ggplot2 3.4.0.
## ℹ Please use `after_stat(density)` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
7 Analisis de correlacion
#Calculamos coeficiente de correlacion de pearson, con Mes como x
with(df_mod, cor.test(Mes, Ventas))##
## Pearson's product-moment correlation
##
## data: Mes and Ventas
## t = -0.46078, df = 44, p-value = 0.6472
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.3525050 0.2255369
## sample estimates:
## cor
## -0.06929832#La correlacion va de -1 a1
# Valores externos indican correlacion mas fuerte
#Pueden ser positiva (suben o bajan juntas) o
#Negativa ( si la otra sube la otra baja y viceversa)8 Modelo de Regresion
modelo_lineal <- lm(Ventas ~ Mes, data = df_mod)
stargazer(modelo_lineal, type = "text")##
## ===============================================
## Dependent variable:
## ---------------------------
## Ventas
## -----------------------------------------------
## Mes -19.934
## (43.261)
##
## Constant 2,413.525**
## (1,167.643)
##
## -----------------------------------------------
## Observations 46
## R2 0.005
## Adjusted R2 -0.018
## Residual Std. Error 3,895.282 (df = 44)
## F Statistic 0.212 (df = 1; 44)
## ===============================================
## Note: *p<0.1; **p<0.05; ***p<0.01#El modelo es y=a + bx
# Constant = a
#Pendiente = b
#Mes = x9 Coeficiente del modelo
#Obtener los coeficientes del modelo, es decir a y b
coeficientes <- coef(modelo_lineal)
intercepto <- coeficientes ["(Intercept)"]
pendiente <- coeficientes["Mes"]
# Crear el texto de la ecuacion
ecuacion <- paste0("y =", round(intercepto, 2),
"+", round(pendiente, 2), "x")
ecuacion## [1] "y =2413.52+-19.93x"10 Prediccion
# Extender la serie temporal para 6 meses
df_futuro <- data.frame(Fecha = seq.Date(from = max(df_mod$Fecha)+ months(1),
by = "month",
length.out = 6),
Mes= seq(from = 49, to = 54, by= 1))
# Generar las predicciones para 6 meses con intervalos de confianza
predicciones <- predict(modelo_lineal,
newdata = df_futuro,
interval = "confidence")
predicciones## fit lwr upr
## 1 1436.769 -1069.750 3943.289
## 2 1416.835 -1167.332 4001.002
## 3 1396.902 -1265.503 4059.306
## 4 1376.968 -1364.215 4118.151
## 5 1357.034 -1463.422 4177.490
## 6 1337.100 -1563.083 4237.28311 Juntar los datos
# Convertir conversiones a un data frame
df_predicciones <- as.data.frame(predicciones)
colnames(df_predicciones) <- c("Ventas", "Bajo", "Alto")
#Unir las predicciones y los intervalos de confianza con las fechas
df_futuro <- cbind(df_futuro, df_predicciones) # cbind une por columnas
# Unir con la base de datos original
df_total <- bind_rows(df_mod, df_futuro)
tail(df_total, 7)## # A tibble: 7 × 5
## Fecha Ventas Mes Bajo Alto
## <date> <dbl> <dbl> <dbl> <dbl>
## 1 2017-12-01 1183. 46 NA NA
## 2 2018-01-01 1437. 49 -1070. 3943.
## 3 2018-02-01 1417. 50 -1167. 4001.
## 4 2018-03-01 1397. 51 -1266. 4059.
## 5 2018-04-01 1377. 52 -1364. 4118.
## 6 2018-05-01 1357. 53 -1463. 4177.
## 7 2018-06-01 1337. 54 -1563. 4237.12 Visualizacion
#Grafica con valores pasados y prediccion
df_total %>%
ggplot(., aes(x = Fecha, y = Ventas)) +
geom_point(dat = df_mod) +
geom_line(data = df_mod) +
geom_smooth(method="lm", formula = y ~ x, color = "blue",
data = df_mod) +
geom_ribbon (data = df_futuro, aes(ymin = Bajo, ymax = Alto),
fill = "lightblue") +
geom_point(data = df_futuro, aes(y= Ventas), color= "red") +
geom_line(data = df_futuro, aes(y=Ventas, color= "red",
linetype = "dashed"))