Train-Test split done and PCA representations

MultivariateAnalysis.R

@@ -7,10 +7,36 @@
 #############
 library(stats) # Calculos
 library(factoextra) # Representaciones
+
+############################# ####################
+# CONJUNTO de ENTRENAMIENTO # # CONJUNTO de TEST #
+############################# ####################
+set.seed(5682)
+library(rsample)
+
+jugadoresRepes <- laLigaPlayersStudy[rowsPIn2C[,1], ]
+jugadoresRepes <- rbind(jugadoresRepes, laLigaPlayersStudy[rowsPIn2C[, 2], ])
+rowsQuit <- c(rowsPIn2C[,1], rowsPIn2C[,2])
+laLigaPlayersSplit <- laLigaPlayersStudy[-rowsQuit, ]
+laLigaPlayersSplit$position.1920 <- laLigaPlayers[-rowsQuit,]$position.1920
+
+laLiga_split <- initial_split(laLigaPlayersSplit, prop = 8/10, strata = "position.1920")
+train <- training(laLiga_split)
+train <- train[,-39]
+train <- rbind(train, jugadoresRepes)
+test <- testing(laLiga_split)
+test <- test[,-39]
+
+plotFor3SeasonsForPos(laLigaPlayers, "position")
+plotFor3SeasonsForPos(train, "position")
+plotFor3SeasonsForPos(test, "position")
+
+
+
 #######################################
 # ANALISIS DE COMPONENTES PRINCIPALES #
 #######################################
-pca_laLiga <- prcomp(laLigaPlayersStudy[,-1], scale = TRUE)
+pca_laLiga <- prcomp(train %>% select(-player), scale = TRUE)
 names(pca_laLiga)
 
 dim(pca_laLiga$rotation)
@@ -19,9 +45,13 @@ dim(pca_laLiga$rotation)
 # SCORES
 head(pca_laLiga$x)[1:5]
 
+# Players' names
+rownames(pca_laLiga$x) <- train$player
 # VARIANZA EXPLICADA POR CADA PC
-pca_laLiga$sdev[1]^2 + pca_laLiga$sdev[2]^2 + pca_laLiga$sdev[3]^2
-# Con 48 PC -> Suma de los 2 primeros = 27.57, Suma de los 3 primeros = 33.925
+(pca_laLiga$sdev[1]^2/sum(pca_laLiga$sdev^2) + pca_laLiga$sdev[2]^2/sum(pca_laLiga$sdev^2) )*100
+# + pca_laLiga$sdev[3]^2
+# Esta mal los comentarios
+# Con 48 PC -> Suma de los 2 primeros = 57.44, Suma de los 3 primeros = 33.925
 # Con 35 PC -> Suma de los 2 primeros = 19.96, Suma de los 3 primeros = 24.15
 
 # REPRESENTACIONES
@@ -29,3 +59,36 @@ fviz_pca_ind(pca_laLiga, geom.ind = "point",
              col.ind = "#FC4E07", 
              axes = c(1, 2), 
              pointsize = 1.5) 
+
+# CREACION del vector POSICIONES 
+colores <- function(vec){
+  # la función rainbow() devuelve un vector que contiene el número de colores distintos
+  col <- rainbow(length(unique(vec)))
+  return(col[as.numeric(as.factor(vec))])
+}
+
+# Observaciones sobre PC1 y PC2
+plot(pca_laLiga$x[,1:2], col = colores(c("GK", "DF", "MF", "FW")), 
+     pch = 19, 
+     xlab = "Z1", 
+     ylab = "Z2")
+legend("topright", legend=c("GK", "DF", "MF", "FW"),
+col = colores(c("GK", "DF", "MF", "FW")),pch=19, cex=0.8)
+
+plot(pca_laLiga$x[,1:2], col = colores(unique(laLigaPlayers$position.1718)), 
+     pch = 19, 
+     xlab = "Z1", 
+     ylab = "Z2")
+legend("topright", legend=unique(laLigaPlayers$position.1718),
+       col = colores(unique(laLigaPlayers$position.1718)),pch=19, cex=0.8)
+table(laLigaPlayers$position.1718)
+fviz_pca_var(pca_laLiga, col.var = "cos2", 
+             geom.var = "arrow", 
+             labelsize = 2, 
+             repel = FALSE)
+
+# BIPLOT
+biplot(pca_laLiga, scale = 0, cex = 0.5, col = c("dodgerblue3", "deeppink3"))
+
+# SELECCIÓN DEL NUMERO DE PC
+fviz_screeplot(pca_laLiga, addlabels = TRUE, ylim = c(0, 40))