Projects/Project6.R at master · iamSarang/Projects · GitHub

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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
#Clear the environment

rm(list=ls(all=TRUE))

setwd("D:/INSOFE/CSE7305c/CUTe 3")

#Read the input data that is given

company_data<-read.csv("train.csv",header = T)

#Read the test data that is given

test_data<-read.csv("test.csv",header=T)

#Use head() and tail() functions to get a feel of the data

head(company_data)

tail(company_data)

head(test_data)

tail(test_data)

#Check the structure of the input data

str(company_data)

str(test_data)

#Check the distribution of the input data using the summary function

summary(company_data)

summary(test_data)

sum(is.na(company_data))

sum(is.na(test_data))

#Removing columns that are not necessary

company_data_mod<-company_data[,-c(1,66)]

test_data_mod<-test_data[-c(1)]

#Impute the data

library(DMwR)
library(caret)

company_data[is.na(company_data)]<-0

company_data_final<-company_data

test_data[is.na(test_data)]<-0

test_data_std<-test_data

company_data_final$target<-as.factor(as.character(company_data_final$target))

company_data_final$targetlabel <- ifelse(company_data_final$target == "0", "g", "h")

# Divide the data into test and validation
set.seed(123)

train_RowIDs = createDataPartition(company_data_final$target,p=0.8,list=F)
train = company_data_final[train_RowIDs,]
validation= company_data_final[-train_RowIDs,]
test<-test_data_std
rm(train_RowIDs)

#Build an ensemble model with xgboost
install.packages("xgboost")
library(xgboost)

train_matrix <- xgb.DMatrix(data = as.matrix(train[, !(names(train) %in% c("target", "targetlabel"))]),
                            label = as.matrix(train[, names(train) %in% "target"]))

validation_matrix <- xgb.DMatrix(data = as.matrix(validation[, !(names(validation) %in% c("target", "targetlabel"))]),
                           label = as.matrix(validation[, names(validation) %in% "target"]))

xgb_model_basic <- xgboost(data = train_matrix, max.depth = 2, eta = 1, nthread = 2, nround = 500, objective = "binary:logistic", verbose = 1, early_stopping_rounds = 10)

xgb.save(xgb_model_basic, "xgb_model_basic")

rm(xgb_model_basic)

xgb_model_basic <- xgb.load("xgb_model_basic")


basic_preds <- predict(xgb_model_basic, validation_matrix)

#Choosing the cut off
basic_preds_labels <- ifelse(basic_preds < 0.5, 0, 1)

library(caret)
result<-confusionMatrix(basic_preds_labels, validation$target)

F1<-result$byClass[7]

params_list <- list("objective" = "binary:logitraw",
                    "eta" = 0.1,
                    "early_stopping_rounds" = 10,
                    "max_depth" = 6,
                    "gamma" = 0.5,
                    "colsample_bytree" = 0.6,
                    "subsample" = 0.65,
                    "eval_metric" = "logloss",
                    "silent" = 1)

xgb_model_with_params <- xgboost(data = train_matrix, params = params_list, nrounds = 500, early_stopping_rounds = 20)

basic_params_preds <- predict(xgb_model_with_params, validation_matrix)

basic_params_preds_labels <- ifelse(basic_params_preds < 0.5, 0, 1)

result_bf_tuning<-confusionMatrix(basic_params_preds_labels, validation$target)

F1_bf_tuning<-result_bf_tuning$byClass[7]

#Variable Importance

variable_importance_matrix <- xgb.importance(feature_names = colnames(train_matrix), model = xgb_model_with_params)

xgb.plot.importance(variable_importance_matrix)

sampling_strategy <- trainControl(method = "repeatedcv", number = 5, repeats = 2, verboseIter = F, allowParallel = T)

param_grid <- expand.grid(.nrounds = 40, .max_depth = c(2, 4, 6), .eta = c(0.1, 0.3),
                          .gamma = c(0.6, 0.5, 0.3), .colsample_bytree = c(0.6, 0.4),
                          .min_child_weight = 1, .subsample = c(0.5, 0.6, 0.9))

xgb_tuned_model <- train(x = train[ , !(names(train) %in% c("targetlabel", "target"))],
                         y = train[ , names(train) %in% c("targetlabel")],
                         method = "xgbTree",
                         trControl = sampling_strategy,
                         tuneGrid = param_grid)

xgb_tuned_model$bestTune

plot(xgb_tuned_model)

tuned_params_preds <- predict(xgb_tuned_model, validation[ , !(names(validation) %in% c("targetlabel", "target"))])

result_Validation<-confusionMatrix(tuned_params_preds, validation$targetlabel)

F1_Validation<-result_Validation$byClass[7]

tuned_params_preds_Test <- predict(xgb_tuned_model, test[ , !(names(test) %in% c("targetlabel", "target"))])

#Using the model built, predict the values for test data

basic_params_preds_labels <- ifelse(tuned_params_preds_Test == "g", 0, 1)

#Write the output file

output<-data.frame(test_data$ID,basic_params_preds_labels)

colnames(output)<-c("ID","prediction")

table(output$prediction)

write.csv(output,file="Samplesubmission2.csv")