-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathlogitModel.R
144 lines (123 loc) · 5.41 KB
/
logitModel.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
rm(list = ls())
library(RcmdrMisc)
library(lubridate)
library(ggplot2)
library(e1071)
library(cluster)
library(FactoMineR)
library(factoextra)
library("rpart")
library("ada")
library(randomForest)
library(dplyr)
library(caret)
library(tidyverse)
library('pscl')
library('IDPmisc')
# Set working directory
setwd("C:/Users/NKARMAKAR/Google Drive/i2D Inrix/All i2D Analysis - Site by Site/Predictive Modeling")
# Reading combined i2d data
data2 <- read.csv(file.choose(),header=TRUE, sep = ",", stringsAsFactors=F)
# Checking for LOS distribution in dataset
barplot(table(data2$LOS))
# Filter unwanted columns
colname <- c('Train','TRIP_ID', 'i2D.Speed..mph.','Engine.Load','Jerk..m.s3.','RPM','Throttle.position','Acc_x_est..m.s2.', 'Total_acc','Acc_y_est..m.s2.','tmc_code','LOS','LOS.ratio', 'TMC_Length','No_int','no_lane','SP_Limit','dr_density', 'tmc_code')
# colname <- c('Acc_x_est (m/s2)', 'Total_acc')
data3 <- data2[colname]
data3[,"Speed_Change_Rate"] <- ave(data3$i2D.Speed..mph., data3$TRIP_ID, data3$tmc_code, FUN=function(x) c(0, (diff(x))))
data3[,"RPM_Change_Rate"] <- ave(data3$RPM, data3$TRIP_ID, data3$tmc_code, FUN=function(x) c(0, (diff(x))))
data3[,"Throttle_Change_Rate"] <- ave(data3$Throttle.position, data3$TRIP_ID, data3$tmc_code, FUN=function(x) c(0, (diff(x))))
data3[sapply(data3, is.infinite)] <- 0
data3[,"R1"] <- data3$Speed_Change_Rate/ data3$RPM_Change_Rate
data3[,"R2"] <- data3$Throttle_Change_Rate/ data3$RPM_Change_Rate
data3 <- NaRV.omit(data3)
R1_percentiles <- as.data.frame(quantile(data3$R1, probs = seq(0, 1, by = 0.025)))
R2_percentiles <- as.data.frame(quantile(data3$R2, probs = seq(0, 1, by = 0.025)))
RPM_Change_Rate_percentiles <- as.data.frame(quantile(data3$RPM_Change_Rate, probs = seq(0, 1, by = 0.025)))
Speed_Change_Rate_percentiles <- as.data.frame(quantile(data3$Speed_Change_Rate, probs = seq(0, 1, by = 0.025)))
Throttle_Change_Rate_percentiles <- as.data.frame(quantile(data3$Throttle_Change_Rate, probs = seq(0, 1, by = 0.025)))
Throttle_Position_percentiles <- as.data.frame(quantile(data3$Throttle.position, probs = seq(0, 1, by = 0.025)))
Engine_Load_percentiles <- as.data.frame(quantile(data3$Engine.Load, probs = seq(0, 1, by = 0.025)))
Jerk_percentiles <- as.data.frame(quantile(data3$Jerk..m.s3., probs = seq(0, 1, by = 0.025)))
ACC_percentiles <- as.data.frame(quantile(data3$Acc_x_est..m.s2., probs = seq(0, 1, by = 0.025)))
plot(RPM_Change_Rate_percentiles)
lower <- 0
upper <- 0.0015
#Function for behavior classification
labeling <- function(R1, R2,load){
if ((R1 < lower || R1 > upper) && (R2 < lower || R2 > upper)&& (load < 20 || load > 60)) {
label <- 1
} else {
label <- 0
}
return (label)
}
labeling2 <- function(x){
if ((x > 50 )) {
label <- 1
} else {
label <- 0
}
return (label)
}
#Creating Labels
data3[,"Label2"]<-mapply(labeling2, data3$Engine.Load)
data3[,"Label"]<-mapply(labeling, data3$R1, data3$R2, data3$Engine.Load)
hist(data3$Label)
hist(data3$Label2)
table <-xtabs(~Label2 + LOS, data = data3)
colPercents(table)
table <-xtabs(~Label2 + tmc_code, data = data3)
Label_by_tmc <- t(as.data.frame(colPercents(table)))
tmc <- rownames(Label_by_tmc)
data_tmc<- as.data.frame(cbind(tmc,Label_by_tmc))
# Creating Traiinng Set
trainset <- data3[data3$Train == 1,]
barplot(table(trainset$LOS))
hist(trainset$Label2)
colname <- c('i2D.Speed..mph.','Engine.Load','RPM','Throttle.position','Acc_x_est..m.s2.', 'Total_acc','Acc_y_est..m.s2.','LOS', 'Label2', 'TMC_Length','No_int','no_lane','SP_Limit','dr_density')
train <- trainset[colname]
train$LOS <- as.factor(train$LOS)
train$Label <- as.factor(train$Label2)
# Creating Test Data
testset <- data3[data3$Train == 0,]
barplot(table(testset$LOS))
hist(testset$Label2)
colname <- c('LOS','TMC_Length','No_int','no_lane','SP_Limit', 'dr_density','Label2')
test <- testset[colname]
test$LOS <- as.factor(test$LOS)
test$Label <- as.factor(test$Label2)
model <- glm (Label2 ~ LOS+TMC_Length+No_int+no_lane+SP_Limit+dr_density, data = train, family = binomial)
summary(model)
anova(model, test="Chisq")
#MacFadden R2 index
pR2(model)
# Prediction
table <-xtabs(~Label2 + LOS, data = test)
table(test$Label2)
colPercents(table)
rowPercents(test$Label2)
totPercents(table(test$Label2))
fitted.results <- predict(model,newdata=subset(test),type='response')
fitted.results <- ifelse(fitted.results > 0.3,1,0)
misClasificError <- mean(fitted.results != test$Label2)
table(fitted.results)
table(test$Label2)
print(paste('Accuracy',1-misClasificError))
varImp(model3)
# Reading combined i2d data
test1 <- read.csv(file.choose(),header=TRUE, sep = ",", stringsAsFactors=F)
colname <- c('LOS','TMC_Length','No_int','no_lane','SP_Limit', 'dr_density')
test2 <- test1[colname]
model2 <- glm (Label2 ~ LOS+TMC_Length+No_int+no_lane+SP_Limit+dr_density, data = data3, family = binomial)
test2.results <- as.data.frame(predict(model2, test2, type = 'response'))
pR2(model2)
model3 <- glm (Label ~ LOS+TMC_Length+No_int+no_lane+SP_Limit+dr_density, data = data3, family = binomial)
test3.results <- as.data.frame(predict(model3, test2, type = 'response'))
summary(model3)
pR2(model3)
# Precision and Recall using Caret Package
precision(tbl_2_1)
precision(data = tbl_2_1_pred, reference = tbl_2_1_truth, relevant = "Relevant")
recall(tbl_2_1)
recall(data = tbl_2_1_pred, reference = tbl_2_1_truth, relevant = "Relevant")