corpo_analyse.Rmd

---
title: "Corpo Analyse"
output: html_document
---

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

# Import et aggrégations de données

```{r}
caracteristiques.2019 <- read.csv2("data/caracteristiques-2019.csv", stringsAsFactors=TRUE)
carac.erronnes <- c(201900033874, 201900035540, 201900035541, 201900007105, 201900058723, 201900017468, 201900005927, 201900033236, 201900006562
                    , 201900056421, 201900000632, 201900057241, 201900048863, 201900008746, 201900056494, 201900047454, 201900021441, 201900025903
                    ,201900020406)
caracteristiques.filtre <- caracteristiques.2019 %>% filter(!Num_Acc %in% carac.erronnes)

lieux.2019 <- read.csv2("data/lieux-2019.csv", sep=";", stringsAsFactors=TRUE)

vehicules.2019 <- read.csv("data/vehicules-2019.csv", sep=";")

usagers.2019 <- read.csv("data/usagers-2019.csv", sep=";")

acc_place_19 <- inner_join(caracteristiques.filtre,lieux.2019,by=c("Num_Acc"="Num_Acc"))

user_car_19<- inner_join(vehicules.2019,usagers.2019,by=c("Num_Acc"="Num_Acc", "id_vehicule" = "id_vehicule", "num_veh" = "num_veh"))

total_19 <- inner_join(acc_place_19,user_car_19,by=c("Num_Acc"="Num_Acc"))

```

## Preprocessing
```{r}

# Approche par les conditions
crit_condition = total_19[c("Num_Acc","lum","agg","int","atm","catr","circ","surf","prof","hrmn")]

# Apporche par l'impact sur véhicule
crit_impact = total_19[c("Num_Acc","col")]

crit_condition <- crit_condition %>% mutate_all(as.factor)
summary(crit_condition)

# Par soucis de simplification on regroupe les luminosités
# Nuit avec éclairage public non allumé --> Nuit sans éclairage public
# Nuit avec éclairage public allumé --> Crépuscule ou aube
levels(crit_condition$lum) <-  factor(c("1", "2", "3", "3", "2"))

# Intersection
levels(crit_condition$int) <-  factor(c("1", "2", "2", "2", "2", "3", "3", "4", "5" ))

# Conditions Atmosphériques
levels(crit_condition$atm) <-  factor(c("1", "1", "1", "2", "3","3", "3", "2", "1", "1"))

# Catégories de routes
levels(crit_condition$catr) <-  factor(c("1", "1", "1", "2", "2", "3", "1", "4"))

# Regime de circulation
levels(crit_condition$circ) <-  factor(c("0", "1", "2", "1", "3"))

# Surface: regroupement pour simplification
# 0 : NR, 1: Normale, 2: Glissant, 3: Eau, 4: Autre
# levels(crit_condition$surf) <-  factor(c("0", "1", "2", "3", "3","2", "3", "2", "2","4"))
levels(crit_condition$surf) <-  factor(c("0", "1", "2", "3", "3","2", "3", "2", "2","4"))

# Profil de la route
levels(crit_condition$prof) <-  factor(c("0", "1", "2", "3", "4"))

```


# Analyses
## ACP
```{r}
summary(crit_condition)
dim(crit_condition)
```

On limite le champs d'analyse à 6 vars

```{r}
data = crit_condition %>% 
  mutate_all(as.character)
data <- data %>% select(-c(1,7,9,10))
d.m <- dummyVars(" ~ .", data = data)
data_ohe <- data.frame(predict(d.m, newdata = data))
```



```{r}
library(FactoMineR)
res.pca <- PCA(data_ohe)
```

En condition de plein jour sur surface normale et en condition atmosphérique normale.
Hors agglomeration, hors intersection et sur routes
En agglomeration, sur voie communales ou hors réseau public

```{r}
barplot(res.pca$eig[,2], 
        names=paste("Dim",1:nrow(res.pca$eig)),
        main="inertie expliquée")
```
```{r}
plot(res.pca,choix="var", axes = c(1,2))
```