-
Notifications
You must be signed in to change notification settings - Fork 0
/
InitialRF.R
143 lines (109 loc) · 4.71 KB
/
InitialRF.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
install.packages("lubridate")
install.packages('rattle')
install.packages('rpart.plot')
install.packages('RColorBrewer')
install.packages('randomForest')
install.packages('dplyr')
install.packages('ggplot2')
install.packages('MLmetrics')
install.packages('caTools')
library(rattle)
library(rpart.plot)
library(RColorBrewer)
library(randomForest) # classification algorithm
library(ggplot2)#Charts
library(dplyr)#Regex
library(lubridate)#For Date Functions
library(rpart)#For CART trees
library(MLmetrics)#For calculating Log Loss
library(caTools)#for splitting training data into training and testing sets
setwd("C:\\Users\\v-arsol\\Documents\\kaggle")
train=read.csv('train.csv')
test=read.csv('test.csv')
summary(train)
names(train)[1] <- 'ID'
test$ID <- as.character(test$ID)
train=bind_rows(train,test)
train$TimeValue <- sapply(as.character(train$AgeuponOutcome),
function(x) strsplit(x, split = " ")[[1]][1])
train$UnitofTime <- sapply(as.character(train$AgeuponOutcome),
function(x) strsplit(x, split = " ")[[1]][2])
train$UnitofTime <- gsub('s', '', train$UnitofTime)
summary(train$UnitofTime)
multiplier <- ifelse(train$UnitofTime == 'day', 1,
ifelse(train$UnitofTime == 'week', 7,
ifelse(train$UnitofTime == 'month', 30, # Close enough
ifelse(train$UnitofTime == 'year', 365, NA))))
# Apply our multiplier
train$TimeValue=as.numeric(train$TimeValue)
train$AgeinDays <- train$TimeValue * multiplier
summary(train$AgeinDays)
#Operations on Name. Creating a flag for animals having a name
train$Name=as.character(train$Name)
train$Name <- ifelse(nchar(train$Name)==0, 'Nameless', train$Name)
train$HasName[train$Name == 'Nameless'] <- 0
train$HasName[train$Name != 'Nameless'] <- 1
#Splitting the date time column
train$year=year(train$DateTime)
train$month=month(train$DateTime)
train$hour=hour(train$DateTime)
train$DayofWeek=weekdays(as.Date(train$DateTime))
train$IsWeekend <- ifelse(train$DayofWeek == 'Saturday', 1,
ifelse(train$DayofWeek == 'Sunday', 1,0))
#Predicting missing values for age
age_fit <- rpart(AgeinDays ~ AnimalType + HasName,
data = train[!is.na(train$AgeinDays), ],
method = 'anova')
train$AgeinDays[is.na(train$AgeinDays)] <- predict(age_fit, train[is.na(train$AgeinDays), ])
train$SexuponOutcome[is.na(train$SexuponOutcome)]="Unknown"
sum(is.na(train$SexuponOutcome))
train$Lifestage[train$AgeinDays < 365] <- 'baby'
train$Lifestage[train$AgeinDays >= 365] <- 'adult'
train$Lifestage[train$AgeinDays >= 5475] <- 'old'
train$Breed <- gsub('Mix', '/', train$Breed)
train$IsMix <- ifelse(grepl(c('/'), train$Breed), 1, 0)
sum(train$IsMix)
train$IsDomestic <- ifelse(grepl(c('Domestic'), train$Breed), 1, 0)
train$NewColor <- sapply(train$Color,
function(x) strsplit(x, split = '/| ')[[1]][1])
train$NewBreed <- sapply(train$Breed,
function(x) strsplit(x, split = '/| ')[[1]][1])
train$Lifestage <- factor(train$Lifestage)
train$SexuponOutcome <- factor(train$SexuponOutcome)
train$IsMix <- factor(train$IsMix)
train$NewColor <- factor(train$NewColor)
train$IsDomestic <- factor(train$IsDomestic)
train$NewBreed <- factor(train$NewBreed)
summary(train$NewColor)
levels(train$NewColor)
levels(train$NewBreed)
test <- train[26730:nrow(train), ]
train <- train[1:26729, ]
set.seed(101)
sample = sample.split(train$OutcomeType, SplitRatio = .75)
train1 = subset(train, sample == TRUE)
train2 = subset(train, sample == FALSE)
#Random Forest
rf_mod <- randomForest(OutcomeType ~ AnimalType+Lifestage+HasName+IsWeekend+hour+SexuponOutcome+NewColor+AgeinDays+IsMix+IsDomestic+month,
data = train1,
ntree = 500,
importance = TRUE)
rf_mod
plot(rf_mod, ylim=c(0,1))
legend('topright', colnames(rf_mod$err.rate), col=1:6, fill=1:6)
importance <- importance(rf_mod)
varImportance <- data.frame(Variables = row.names(importance),
Importance = round(importance[ ,'MeanDecreaseGini'],2))
varImportance
# Create a rank variable based on importance
rankImportance <- varImportance %>%
mutate(Rank = paste0('#',dense_rank(desc(Importance))))
plot(rf_mod)
varImpPlot(rf_mod)
#testing the model on remaining training set and calculating log loss
prediction <- predict(rf_mod, train2, type = 'vote')
MultiLogLoss(train2$OutcomeType,prediction)
#Applying the model on the test set
prediction <- predict(rf_mod, test, type = 'vote')
solution <- data.frame('ID' = test$ID, prediction)
write.csv(solution, 'rf_solution.csv', row.names = F)