projetS5_Imane.Rmd

---
title: "Projet Transdev"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r}
library(plyr)
library(dplyr)
library(magrittr)
library(lubridate)
library(ggplot2)
library(tidyverse)
library(sf)
library(caret)
library(class)
library(factoextra)
library(ggmap)
library(osmdata)
library(geosphere)
library(grid)
library(lattice)
library(modeltools)
library(stats4)
library(flexclust)
library(keras)
```

# 2°) Clustering des lignes 

## 2.1°) Préparation des données

```{r echo=FALSE}
dfTypeJour1_ligne = subset.data.frame(df,df$TYPEJOUR == "Dimanche" | df$FERIE == 1)[,c("LIGNE","HORAIRE")]
dfTypeJour1_ligne$HORAIRE = hour(dfTypeJour1_ligne$HORAIRE)
dfTypeJour1_ligne = dfTypeJour1_ligne %>% count(c("LIGNE", "HORAIRE"))
```

```{r echo=FALSE}
dfTypeJour2_ligne = subset.data.frame(df,df$TYPEJOUR == "Samedi" & df$FERIE == 0)[,c("LIGNE","HORAIRE")]
dfTypeJour2_ligne$HORAIRE = hour(dfTypeJour2_ligne$HORAIRE)
dfTypeJour2_ligne = dfTypeJour2_ligne %>% count(c("LIGNE", "HORAIRE"))
```

```{r echo=FALSE}
dfTypeJour3_ligne = subset.data.frame(df,df$TYPEJOUR %in% c("Lundi","Mardi","Mercredi","Jeudi","Vendredi") & df$FERIE == 0 & df$SCOLAIRE == 1)[,c("LIGNE","HORAIRE")]
dfTypeJour3_ligne$HORAIRE = hour(dfTypeJour3_ligne$HORAIRE)
dfTypeJour3_ligne = dfTypeJour3_ligne %>% count(c("LIGNE", "HORAIRE"))
```

```{r echo=FALSE}
dfTypeJour4_ligne = subset.data.frame(df,df$TYPEJOUR %in% c("Lundi","Mardi","Mercredi","Jeudi","Vendredi") & df$FERIE == 0 & df$SCOLAIRE == 0)[,c("LIGNE","HORAIRE")]
dfTypeJour4_ligne$HORAIRE = hour(dfTypeJour4_ligne$HORAIRE)
dfTypeJour4_ligne = dfTypeJour4_ligne %>% count(c("LIGNE", "HORAIRE"))
```

```{r echo=FALSE}
labels_ligne = StatsLignes$Var1
heures = c(0,1,seq(5,23))
```

```{r echo=FALSE}
df1_ligne = data.frame(matrix(0, ncol = 24, nrow = 53))
rownames(df1_ligne) <- labels_ligne
for (k in 1:length(dfTypeJour1_ligne[,1])){
  df1_ligne[c(dfTypeJour1_ligne$LIGNE[k]),dfTypeJour1_ligne$HORAIRE[k]+1]=dfTypeJour1_ligne$freq[k]
}
dfTypeJour1_ligne = df1_ligne
df1_ligne <- echelle(df1_ligne)
df1_ligne = df1_ligne[ , colSums(is.na(df1_ligne)) < nrow(df1_ligne)]
df1_ligne <- na.omit(df1_ligne)
```

```{r echo=FALSE}
df2_ligne = data.frame(matrix(0, ncol = 24, nrow = 53))
rownames(df2_ligne) <- labels_ligne
for (k in 1:length(dfTypeJour2_ligne[,1])){
  df2_ligne[c(dfTypeJour2_ligne$LIGNE[k]),dfTypeJour2_ligne$HORAIRE[k]+1]=dfTypeJour2_ligne$freq[k]
}
dfTypeJour2_ligne = df2_ligne
df2_ligne <- echelle(df2_ligne)
df2_ligne = df2_ligne[ , colSums(is.na(df2_ligne)) < nrow(df2_ligne)]
df2_ligne <- na.omit(df2_ligne)
```

```{r echo=FALSE}
df3_ligne = data.frame(matrix(0, ncol = 24, nrow = 53))
rownames(df3_ligne) <- labels_ligne
for (k in 1:length(dfTypeJour3_ligne[,1])){
  df3_ligne[c(dfTypeJour3_ligne$LIGNE[k]),dfTypeJour3_ligne$HORAIRE[k]+1]=dfTypeJour3_ligne$freq[k]
}
dfTypeJour3_ligne = df3_ligne
df3_ligne <- echelle(df3_ligne)
df3_ligne = df3_ligne[ , colSums(is.na(df3_ligne)) < nrow(df3_ligne)]
df3_ligne <- na.omit(df3_ligne)
df3_ligne["70",] <- df3_ligne[54,]
df3_ligne["180",] <- df3_ligne[55,]
df3_ligne["280",] <- df3_ligne[56,]
df3_ligne <- df3_ligne[-54:-56,]
```

```{r echo=FALSE}
df4_ligne = data.frame(matrix(0, ncol = 24, nrow = 53))
rownames(df4_ligne) <- labels_ligne
for (k in 1:length(dfTypeJour4_ligne[,1])){
  df4_ligne[c(dfTypeJour4_ligne$LIGNE[k]),dfTypeJour4_ligne$HORAIRE[k]+1]=dfTypeJour4_ligne$freq[k]
}
dfTypeJour4_ligne = df4_ligne
df4_ligne <- echelle(df4_ligne)
df4_ligne = df4_ligne[ , colSums(is.na(df4_ligne)) < nrow(df4_ligne)]
df4_ligne <- na.omit(df4_ligne)
df4_ligne["70",] <- df4_ligne[54,]
df4_ligne["180",] <- df4_ligne[55,]
df4_ligne <- df4_ligne[-54:-55,]
```

```{r echo=FALSE}
dataAClusterParLigne = cbind(df1_ligne,df2_ligne,df3_ligne,df4_ligne)
```

## 2.2) Classification Ascendante Hiérarchique 

```{r}
# calcul des distances euclidiennes entre les points 
lignes.dist <- dist(dataAClusterParLigne)
```


```{r}
cah.ward <- hclust(lignes.dist, method="ward.D")
print(cah.ward)
```
```{r}
plot(cah.ward)
```
```{r}
groupes.cah <- cutree(cah.ward,k=4)
```

Affichage du dendrogramme pour un nb de clusters choisi
```{r}
# remplacer 4 par le nb de clusters choisi
nbClasses=4
source("http://addictedtor.free.fr/packages/A2R/lastVersion/R/code.R")
ordreClasses <- unique(groupes.cah[cah.ward$order])
cPal <- c(1:nbClasses)[ordreClasses]

A2Rplot(x = cah.ward,
        k = nbClasses,
        boxes = FALSE,
        col.up = "gray50",
        col.down = cPal,
        show.labels = TRUE,
        
        main = "Dendrogramme")

legend(x="topleft",
       paste("Cluster",
             1:nbClasses,
             sep=" "),
       cex=1,
       seg.len=4,
       col=cPal,
       pch=NA,
       lty=1,
lwd=4)
```
```{r}
groupes.cah[cah.ward$order]
```


```{r}
#cluster1=list()
#for (k in 1:length( groupes.cah )){
#  if ( unname(groupes.cah[k]) == 1 )
#    cluster1 <-  c(cluster1, names( groupes.cah[k]  ))
#}
#print(cluster1)
```


## 2.2) Représentations des clusters

Représentation des trajets/h en moyenne pour chaque cluster suivant le type de jour (GR1 de 0 à 23, GR2 de 24 à 48...) :
```{r}
dfTousTypesJours_ligne = cbind(dfTypeJour1_ligne,dfTypeJour2_ligne,dfTypeJour3_ligne,dfTypeJour4_ligne)
dfTousTypesJours_ligne <- na.omit(dfTousTypesJours_ligne)
```

Affichage pour 4 clusters
```{r}
new_df <-  t ( data.frame( cluster1= rep(0, times=96) , cluster2= rep(0, times=96), cluster3= rep(0, times=96), cluster4=rep(0, times=96) ) ) 

s1 <- 0
s2 <- 0
s3 <- 0
s4 <- 0

for (i in 1:53) 
  if ( unname(groupes.cah)[i]==1 ){
    s1 <- s1 + 1
    for (k in 1:96)
      new_df[1,k] = new_df[1,k] + dfTousTypesJours_ligne[i,k]
  } else if ( unname(groupes.cah)[i]==2 ) {
    s2 <- s2 + 1
    for (k in 1:96)
      new_df[2,k] = new_df[2,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==3 ){
    s3 <- s3 + 1
    for (k in 1:96)
      new_df[3,k] = new_df[3,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==4 ){
    s4 <- s4 + 1
    for (k in 1:96)
      new_df[4,k] = new_df[4,k] + dfTousTypesJours_ligne[i,k]
 }

for (i in 1:53) 
  new_df[1,i] = new_df[1,i] / s1
  new_df[2,i] = new_df[2,i] / s2
  new_df[3,i] = new_df[3,i] / s3
  new_df[4,i] = new_df[4,i] / s4
```

```{r}
new_df3 <-  data.frame( nb= rep(0, times=384) , cluster= rep(0, times=384), heure = rep(seq(1:96), times=4)) 

for (i in 1:96){
  new_df3[i,1]  <-   new_df["cluster1",i]
  new_df3[i,2]  =   1
}
for (i in 97:192){
  new_df3[i,1]  <-   new_df["cluster2",i-96]
  new_df3[i,2]  <-   2
}
for (i in 193:288){
  new_df3[i,1]  <-   new_df["cluster3",i-192]
  new_df3[i,2]  <-   3
}
for (i in 289:384){
  new_df3[i,1]  <-   new_df["cluster4",i-288]
  new_df3[i,2]  <-   4
}

ggplot(new_df3, aes(x= heure, y= nb, colour=cluster ))  + ggtitle("Moyennes de trajets/h suivant les clusters") + geom_point()
```
Affichage pour 2 clusters
```{r}
new_df <-  t ( data.frame( cluster1= rep(0, times=96) , cluster2= rep(0, times=96) ) )

s1 <- 0
s2 <- 0

for (i in 1:53) 
  if ( unname(groupes.cah)[i]==1 ){
    s1 <- s1 + 1
    for (k in 1:96)
      new_df[1,k] = new_df[1,k] + dfTousTypesJours_ligne[i,k]
  } else if ( unname(groupes.cah)[i]==2 ) {
    s2 <- s2 + 1
    for (k in 1:96)
      new_df[2,k] = new_df[2,k] + dfTousTypesJours_ligne[i,k]
 } 

for (i in 1:53) 
  new_df[1,i] = new_df[1,i] / s1
  new_df[2,i] = new_df[2,i] / s2
```

```{r}
new_df3 <-  data.frame( nb= rep(0, times=192) , cluster= rep(0, times=192), heure = rep(seq(1:96), times=2)) 

for (i in 1:96){
  new_df3[i,1]  <-   new_df["cluster1",i]
  new_df3[i,2]  =   1
}
for (i in 97:192){
  new_df3[i,1]  <-   new_df["cluster2",i-96]
  new_df3[i,2]  <-   2
}

ggplot(new_df3, aes(x= heure, y= nb, colour=cluster ))  + ggtitle("Moyennes de trajets/h suivant les clusters") + geom_point()
```


Affichage pour 6 clusters
```{r}
new_df <-  t ( data.frame( cluster1= rep(0, times=96) , cluster2= rep(0, times=96), cluster3= rep(0, times=96), cluster4=rep(0, times=96), cluster5= rep(0, times=96), cluster6=rep(0, times=96) ) )

s1 <- 0
s2 <- 0
s3 <- 0
s4 <- 0
s5 <- 0
s6 <- 0

for (i in 1:53) 
  if ( unname(groupes.cah)[i]==1 ){
    s1 <- s1 + 1
    for (k in 1:96)
      new_df[1,k] = new_df[1,k] + dfTousTypesJours_ligne[i,k]
  } else if ( unname(groupes.cah)[i]==2 ) {
    s2 <- s2 + 1
    for (k in 1:96)
      new_df[2,k] = new_df[2,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==3 ){
    s3 <- s3 + 1
    for (k in 1:96)
      new_df[3,k] = new_df[3,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==4 ){
    s4 <- s4 + 1
    for (k in 1:96)
      new_df[4,k] = new_df[4,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==5 ){
    s5 <- s5 + 1
    for (k in 1:96)
      new_df[5,k] = new_df[5,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==6 ){
    s6 <- s6 + 1
    for (k in 1:96)
      new_df[6,k] = new_df[6,k] + dfTousTypesJours_ligne[i,k]
 }

for (i in 1:53) 
  new_df[1,i] = new_df[1,i] / s1
  new_df[2,i] = new_df[2,i] / s2
  new_df[3,i] = new_df[3,i] / s3
  new_df[4,i] = new_df[4,i] / s4
  new_df[5,i] = new_df[5,i] / s5
  new_df[6,i] = new_df[6,i] / s6
```

```{r}
new_df3 <-  data.frame( nb= rep(0, times=576) , cluster= rep(0, times=576), heure = rep(seq(1:96), times=6)) 

for (i in 1:96){
  new_df3[i,1]  <-   new_df["cluster1",i]
  new_df3[i,2]  =   1
}
for (i in 97:192){
  new_df3[i,1]  <-   new_df["cluster2",i-96]
  new_df3[i,2]  <-   2
}
for (i in 193:288){
  new_df3[i,1]  <-   new_df["cluster3",i-192]
  new_df3[i,2]  <-   3
}
for (i in 289:384){
  new_df3[i,1]  <-   new_df["cluster4",i-288]
  new_df3[i,2]  <-   4
}

for (i in 385:480){
  new_df3[i,1]  <-   new_df["cluster3",i-384]
  new_df3[i,2]  <-   5
}
for (i in 481:576){
  new_df3[i,1]  <-   new_df["cluster4",i-480]
  new_df3[i,2]  <-   6
}

ggplot(new_df3, aes(x= heure, y= nb, colour=cluster ))  + ggtitle("Moyennes de trajets/h suivant les clusters") + geom_point()
```

Affichage pour 3 clusters
```{r}
new_df <-  t ( data.frame( cluster1= rep(0, times=96) , cluster2= rep(0, times=96), cluster3= rep(0, times=96) )) 

s1 <- 0
s2 <- 0
s3 <- 0

for (i in 1:53) 
  if ( unname(groupes.cah)[i]==1 ){
    s1 <- s1 + 1
    for (k in 1:96)
      new_df[1,k] = new_df[1,k] + dfTousTypesJours_ligne[i,k]
  } else if ( unname(groupes.cah)[i]==2 ) {
    s2 <- s2 + 1
    for (k in 1:96)
      new_df[2,k] = new_df[2,k] + dfTousTypesJours_ligne[i,k]
 } else if ( unname(groupes.cah)[i]==3 ){
    s3 <- s3 + 1
    for (k in 1:96)
      new_df[3,k] = new_df[3,k] + dfTousTypesJours_ligne[i,k]
 } 
for (i in 1:53) 
  new_df[1,i] = new_df[1,i] / s1
  new_df[2,i] = new_df[2,i] / s2
  new_df[3,i] = new_df[3,i] / s3
```

```{r}
new_df3 <-  data.frame( nb= rep(0, times=288) , cluster= rep(0, times=288), heure = rep(seq(1:96), times=3)) 

for (i in 1:96){
  new_df3[i,1]  <-   new_df["cluster1",i]
  new_df3[i,2]  =   1
}
for (i in 97:192){
  new_df3[i,1]  <-   new_df["cluster2",i-96]
  new_df3[i,2]  <-   2
}
for (i in 193:288){
  new_df3[i,1]  <-   new_df["cluster3",i-192]
  new_df3[i,2]  <-   3
}

ggplot(new_df3, aes(x= heure, y= nb, colour=cluster ))  + ggtitle("Moyennes de trajets/h suivant les clusters") + geom_point()
```