3D PCA and Var Biplot

MultivariateAnalysis.R

@@ -6,7 +6,8 @@
 # LIBRERIAS #
 #############
 library(stats) # Calculos
-library(factoextra) # Representaciones
+library(factoextra) # PCAs representations
+library(plotly) # 3D PCA representations
 
 ############################# ####################
 # CONJUNTO de ENTRENAMIENTO # # CONJUNTO de TEST #
@@ -17,16 +18,18 @@ 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
+laLigaPlayersSplit$position <- laLigaPlayers[-rowsQuit,]$position.1920
 
-laLiga_split <- initial_split(laLigaPlayersSplit, prop = 8/10, strata = "position.1920")
+laLiga_split <- initial_split(laLigaPlayersSplit, prop = 8/10, strata = "position")
 train <- training(laLiga_split)
 train <- train[,-39]
 train <- rbind(train, jugadoresRepes)
 test <- testing(laLiga_split)
 test <- test[,-39]
 
+# ToDo
 plotFor3SeasonsForPos(laLigaPlayers, "position")
 plotFor3SeasonsForPos(train, "position")
 plotFor3SeasonsForPos(test, "position")
@@ -54,12 +57,25 @@ rownames(pca_laLiga$x) <- train$player
 # 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
+
+repeatedPlayers <- which(duplicated(train$player) == TRUE)
+for (player in repeatedPlayers){
+  train$player[player] <- paste(train$player[player], 2, sep = "")
+}
+rownames(pca_laLiga$x) <- train$player
+
+# REPRESENTATIONS
+# 2D
 fviz_pca_ind(pca_laLiga, geom.ind = "point", 
-             col.ind = "#FC4E07", 
+             col.ind = "aquamarine3", 
              axes = c(1, 2), 
              pointsize = 1.5) 
+# 3D 
 
+data <- data.frame(pca_laLiga$x)
+fig <- plot_ly(data, x = ~PC1, y = ~PC2, z = ~PC3, colors = c('aquamarine3') ) %>%
+  add_markers(size = 12)
+fig
 # CREACION del vector POSICIONES 
 colores <- function(vec){
   # la función rainbow() devuelve un vector que contiene el número de colores distintos
@@ -82,13 +98,17 @@ plot(pca_laLiga$x[,1:2], col = colores(unique(laLigaPlayers$position.1718)),
 legend("topright", legend=unique(laLigaPlayers$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))
+# BIPLOT (toSave)
+biplot(pca_laLiga, scale = 0, cex = 0.5, col = c("khaki4", "darkorchid3"))
+fviz_pca_biplot(pca_laLiga)
+fviz_pca_var(pca_laLiga, 
+             col.var = "contrib", 
+             gradient.cols = c("#FC4E07", "#E7B800", "#006600"),
+             labelsize = 3, 
+             repel = TRUE)
+
+# SELECCIÓN DEL NUMERO DE PC (700x550)
+fviz_screeplot(pca_laLiga, addlabels = TRUE, ylim = c(0, 40), 
+               xlab = "Dimensiones", ylab = "Porcentaje de variabilidad explicada")