CAH file

03_CAH_clust.R

@@ -0,0 +1,54 @@
+library(data.table)
+library(mltools)
+library(dplyr)
+library(tidyr)
+library(here)
+library(ggplot2)
+library(ggdendro)
+library(factoextra)
+
+data <- data.frame(read.csv(file.path(here(), "data", "02_preprocessed_data.csv"),
+    header = TRUE
+), stringsAsFactors = TRUE)
+
+colnames(data)[5:ncol(data)] <- c(
+    "1AC-OUVERTURE", "1AC-MI",
+    "2A-UEchoix-S1-UC1", "2A-UEchoix-S1-UC2", "2A-UEchoix-S2-UC3",
+    "2A-UEchoix-S2-UC4",
+    "2A-UEchoix-S2-UC5", "2A-UEchoix-S2-UC6",
+    "2A-Projet-S2"
+)
+
+data <- data %>%
+    mutate_if(sapply(data, is.character), as.factor)
+
+selected_cols <- c(
+    "parcours", "domaine2A", "1AC-MI",
+    "2A-UEchoix-S1-UC1", "2A-UEchoix-S1-UC2", "2A-UEchoix-S2-UC4",
+    "2A-UEchoix-S2-UC3", "2A-UEchoix-S2-UC5", "2A-UEchoix-S2-UC6",
+    "2A-Projet-S2", "1AC-OUVERTURE"
+)
+
+onehot_data <- one_hot(as.data.table(data), cols = selected_cols, sparsifyNAs = TRUE)
+
+#  Fonctionne bien avec binary
+dist_eucl <- dist(x = onehot_data[, -c(1, 2, 3)], method = "binary")
+
+hclust_avg <- hclust(dist_eucl, method = "average")
+
+dhc <- as.dendrogram(hclust_avg)
+
+plotdata <- dendro_data(dhc, type = "rectangle")
+p <- ggplot(segment(plotdata)) +
+    geom_segment(aes(x = x, y = y, xend = xend, yend = yend)) +
+    coord_flip() +
+    scale_y_reverse(expand = c(0.2, 0))
+p + theme_dendro()
+
+fviz_nbclust(onehot_data[, -c(1,2,3)], FUNcluster = hcut, k.max = 30)
+
+cut_avg <- cutree(hclust_avg, k = 6)
+names(cut_avg) <- data[["ine"]]
+table(cut_avg)
+
+onehot_data[["cluster"]] <- cut_avg