89 lines
2.5 KiB
R
89 lines
2.5 KiB
R
library(data.table)
|
||
library(mltools)
|
||
library(dplyr)
|
||
library(tidyr)
|
||
library(here)
|
||
library(ggplot2)
|
||
library(ggdendro)
|
||
library(factoextra)
|
||
|
||
set.seed(1234)
|
||
data <- data.frame(read.csv(file.path(here(), "data", "02_preprocessed_data.csv"),
|
||
header = TRUE
|
||
), stringsAsFactors = TRUE)
|
||
|
||
data_folder_path <- file.path(here(), "data", "04_CAH_par_dominante")
|
||
|
||
if (!dir.exists(data_folder_path)) {
|
||
dir.create(data_folder_path)
|
||
}
|
||
|
||
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(
|
||
"dominante3A", "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"
|
||
)
|
||
|
||
seuil_effectif <- 20L
|
||
|
||
# On récupère les dominantes supérieures au seuil
|
||
vec_dominantes <- na.omit(sapply(unique(data[["dominante3A"]]), function(dominante) {
|
||
ifelse(sum(data[["dominante3A"]] == dominante) > seuil_effectif,
|
||
as.character(dominante), NA
|
||
)
|
||
}))
|
||
|
||
for (dominante in vec_dominantes) {
|
||
message("Dominante ", dominante)
|
||
data_dominante <- data %>% filter(dominante3A == dominante)
|
||
onehot_data <- one_hot(as.data.table(data_dominante), cols = selected_cols, sparsifyNAs = TRUE)
|
||
|
||
# Fonctionne bien avec binary
|
||
dist <- dist(x = onehot_data[, -1], method = "binary")
|
||
|
||
hclust_avg <- hclust(dist, 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)) +
|
||
ggtitle(paste0("Dominante : ", dominante))
|
||
p + theme_dendro()
|
||
|
||
print(fviz_nbclust(onehot_data,
|
||
FUNcluster = hcut
|
||
)+
|
||
ggtitle(paste0("Dominante : ", dominante)))
|
||
|
||
nb_clust_dominante <- as.integer(readline(prompt = paste0(
|
||
"Nb clusters ",
|
||
dominante,
|
||
" : "
|
||
)))
|
||
|
||
cut_avg <- cutree(hclust_avg, k = nb_clust_dominante)
|
||
names(cut_avg) <- data_dominante[["ine"]]
|
||
table(cut_avg)
|
||
|
||
data_dominante[["cluster"]] <- cut_avg
|
||
|
||
write.csv(data_dominante, file.path(
|
||
data_folder_path,
|
||
paste0("04_", dominante, "_cah_results.csv")
|
||
), row.names = FALSE)
|
||
}
|