-
Notifications
You must be signed in to change notification settings - Fork 0
/
week15_assi7_sol1.R
408 lines (325 loc) · 14.2 KB
/
week15_assi7_sol1.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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
# # -*- coding: utf-8 -*-
# """Week15 Assi7 Sol1.ipynb
#
# Automatically generated by Colaboratory.
#
# Original file is located at
# https://colab.research.google.com/drive/1_Et_v9mmzBrmbE8jR1XTGHR_m1d5DieY
# """
###########################################################################
## Week-15, Homework-7, Sol-1
## Sreya Dhar
## Created: Dec 09, 2020
## Edited: Dec 16, 2020
###########################################################################
rm(list=ls())
## installing all the libraries in R kernel
# install.packages("corrplot")
# install.packages("forecast")
# install.packages("zoo")
# install.packages("rsample")
# install.packages("leaps")
# install.packages("car")
# install.packages("caret")
# install.packages("ROCR")
# install.packages("PerformanceAnalytics")
# install.packages("funModeling")
# install.packages("hrbrthemes")
# install.packages("ggthemes")
# install.packages("GGally")
# install.packages("glmnet")
# install.packages("ISLR")
# install.packages("kableExtra")
# install.packages("broom")
# install.packages("knitr")
# install.packages("psych")
# install.packages("aod")
# install.packages("epiDisplay")
# install.packages("e1071")
# install.packages("class")
# install.packages("rpart.plot")
# install.packages("party")
# install.packages("partykit")
# install.packages("rattle")
## importing the libraries in R kernel
library(ggplot2)
library(dplyr)
library(tidyverse)
library(tidyr)
library(corrplot)
library(repr)
library(reshape2)
library(forecast)
library(zoo)
library(rsample)
library(gplots)
library(ROCR)
library(class)
library(readr)
library(leaps)
library(car)
library(PerformanceAnalytics)
library(funModeling)
library(gridExtra)
library(caret)
library(MASS)
library(Hmisc)
library(hrbrthemes)
library(GGally)
library(glmnet)
library(pROC)
library(psych)
library(aod)
library(epiDisplay)
library(e1071)
library(ggthemes)
library(kableExtra)
library(broom)
library(knitr)
library(devtools)
library(rpart) #for trees
library(rattle) # Fancy tree plot This is a difficult library to install (https://gist.github.com/zhiyzuo/a489ffdcc5da87f28f8589a55aa206dd)
library(rpart.plot) # Enhanced tree plots
library(RColorBrewer) # Color selection for fancy tree plot
library(party) # Alternative decision tree algorithm
library(partykit) # Convert rpart object to BinaryTree
library(randomForest)
library(viridis)
library(tree)
library(factoextra)
library(neuralnet)
## set directory ##
setwd("C:/File G/EAS 506 Statistical Mining I/Week 15/Assignment-7")
## converting the RData file into .csv file
load("C:/File G/EAS 506 Statistical Mining I/Week 15/Assignment-7/cleveland.RData")
write.csv(cleveland,'cleveland.csv')
cleveland <- read.csv("cleveland.csv", header = TRUE)
cleveland<- cleveland[,-1]
cleveland<- cleveland[,-15]
glimpse(cleveland)
names(cleveland)
status(cleveland)
profiling_num(cleveland)
describe(cleveland)
summary(cleveland)
options(repr.plot.width=6, repr.plot.height=6, repr.plot.res = 200)
cleveland_n <- cleveland %>% mutate_if(is.character, as.factor)
cleveland_n1 <- cleveland_n %>% mutate_if(is.factor, as.numeric)
L <- cor(cleveland_n1)
corrplot(L, method = "circle", type = "lower")
pairs.panels(cleveland_n1[,1:13], main = "Pairs plot on Cleveland dataset, unclassed on Class", pch = 21, bg = c("blue", "green")[unclass(cleveland_n1$diag1)], hist.col="red")
plot_num(cleveland_n1)
################## Preparing the data for Analysis ########################
# ## min-max scaling on vehicle dataset predictors prior to classification
normalized<-function(y) {
x<-y[!is.na(y)]
x<-(x - min(x)) / (max(x) - min(x))
y[!is.na(y)]<-x
return(y)
}
cleveland[,1:13]<-apply(cleveland_n1[,c(1:13)],2,normalized)
head(cleveland)
#####################################################
####################### ANN #########################
#####################################################
data <- cleveland %>% mutate_if(is.character, as.factor)
data_1 <- data %>% mutate_if(is.factor, as.numeric)
data_1$diag1<- data_1$diag1 -1
set.seed(4444) ## seeding the sampling
data_split <- initial_split(data_1, prop = 0.75) ## splitting the data by library 'rsample'
data_train <- training(data_split)
data_test <- testing(data_split)
nn1=neuralnet(diag1 ~., data=data_train,
hidden=c(7, 3),
act.fct = "logistic",
linear.output = FALSE,
err.fct = 'ce',
likelihood = TRUE)
# plot neural network
plot(nn1, rep = 'best')
train_err <- nn1$result.matrix[1,1]
paste("CE Error: ", round(train_err, 3))
nn1_AIC <- nn1$result.matrix[4,1]
paste("AIC: ", round(nn1_AIC,3))
nn1_BIC <- nn1$result.matrix[5,1]
paste("BIC: ", round(nn1_BIC, 3))
################# Classification Hyperparameters #######################
########################################################################
### 2-Hidden Layers, Layer-1 2-neurons, Layer-2, 1-neuron
nn2 <- neuralnet(diag1 ~., data=data_train,
linear.output = FALSE,
err.fct = 'ce',
likelihood = TRUE,
hidden = c(8,4))
# 2-Hidden Layers, Layer-1 2-neurons, Layer-2, 2-neurons
nn3 <- neuralnet(diag1 ~., data = data_train,
linear.output = FALSE,
err.fct = 'ce',
likelihood = TRUE,
hidden = c(7))
# 2-Hidden Layers, Layer-1 1-neuron, Layer-2, 2-neuron
nn4 <- neuralnet(diag1 ~., data = data_train,
linear.output = FALSE,
err.fct = 'ce',
likelihood = TRUE,
hidden = c(9,5))
# Bar plot of results
nn_ics <- tibble('Network' = rep(c("NN1", "NN2", "NN3", "NN4"), each = 3),
'Metric' = rep(c('AIC', 'BIC', 'ce Error * 100'), length.out = 12),
'Value' = c(nn1$result.matrix[4,1], nn1$result.matrix[5,1], 100*nn1$result.matrix[1,1],
nn2$result.matrix[4,1], nn2$result.matrix[5,1], 100*nn2$result.matrix[1,1],
nn3$result.matrix[4,1], nn3$result.matrix[5,1], 100*nn3$result.matrix[1,1],
nn4$result.matrix[4,1], nn4$result.matrix[5,1], 100*nn4$result.matrix[1,1]))
nn_ics %>%
ggplot(aes(Network, Value, fill = Metric)) +
geom_col(position = 'dodge') +
ggtitle("AIC, BIC, and Cross-Entropy Error of the Classification ANNs", "Note: ce Error displayed is 100 times its true value")
nn_3_pred <- predict(nn3, data_train)
nn_3_pred_te <- predict(nn3, data_test)
confusionMatrix(table(actual = data_train$diag1, prediction = round(nn_3_pred)))
confusionMatrix(table(actual = data_test$diag1, prediction = round(nn_3_pred_te)))
results <- data.frame(actual = data_test$diag1, prediction = round(nn_3_pred_te))
###########################################
## put in a loop for tuning.
###########################################
train_err_store <- c()
test_err_store <- c()
for (i in 1:10){
# fit neural network with "i" neurons
nn0 <- neuralnet(diag1 ~., data = data_train,
hidden = i, stepmax = 10^9, err.fct = "ce", linear.output = FALSE)
# calculate the train error
pred <- predict(nn0, newdata = data_train)
y_hat_train <- round(pred)
train_err <- length(which(data_train$diag1 != y_hat_train))/length(y_hat_train)
train_err_store <- c(train_err_store, train_err) #store the error at each iteration
pred <- predict(nn0, newdata = data_test)
y_hat_test <- round(pred)
test_err <- length(which(data_test$diag1 != y_hat_test))/length(y_hat_test)
test_err_store <- c(test_err_store, test_err) #store the error at each iteration
}
train_err_store
test_err_store
err_score <- data.frame(index=c( 1,2,3, 4, 5, 6,7,8, 9, 10),train_err_store, test_err_store)
err_score <- melt(err_score , id.vars = 'index', variable.name = 'series')
#create line plot for each column in data frame
ggplot(err_score, aes(index, value)) +
geom_line(aes(colour = series))+labs(x = "No of Hidden Layers", y= 'Misclassification error rate', title='Tuning hidden Layers of ANN')+theme_bw()
###########################################################################
# #Test the resulting output from best tuned ANN with #hidden input = 6 ###
###########################################################################
nn_best <- neuralnet(diag1 ~., data = data_train,
linear.output = FALSE,
err.fct = 'ce',
likelihood = TRUE,
hidden = c(2))
plot(nn_best, rep = 'best')
nn_best_pred <- predict(nn_best, data_train)
nn_best_pred_te <- predict(nn_best, data_test)
confusionMatrix(table(actual = data_train$diag1, prediction = round(nn_best_pred)))
confusionMatrix(table(actual = data_test$diag1, prediction = round(nn_best_pred_te)))
results <- data.frame(actual = data_test$diag1, prediction = round(nn_best_pred_te))
################################################
pROC_obj <- roc(results[,1],results[,2],
smoothed = TRUE,
# arguments for ci
ci=TRUE, ci.alpha=0.95, stratified=FALSE,
# arguments for plot
plot=TRUE, auc.polygon=TRUE, max.auc.polygon=TRUE, grid=TRUE,
print.auc=TRUE, show.thres=TRUE)
sens.ci <- ci.se(pROC_obj)
plot(sens.ci, type="shape", col="lightblue")
## splitting the dataset into train and test sets
set.seed(4444) ## seeding the sampling
data_split <- initial_split(cleveland, prop = 0.75) ## splitting the data by library 'rsample'
data_train <- training(data_split)
data_test <- testing(data_split)
#########################################################
########################### CART ########################
#########################################################
tree_mod = tree(factor(diag1) ~., data = data_train, split='deviance')
plot(tree_mod)
text(tree_mod, cex=1.0)
tree_pred_tr = predict(tree_mod, data_train[,-14], type='class')
tree_tab_tr <- table(Original=data_train$diag1, Prediction=tree_pred_tr)
confusionMatrix(tree_tab_tr)
tree_pred_te = predict(tree_mod, data_test[,-14], type='class')
tree_tab_te <- table(Original=data_test$diag1, Prediction=tree_pred_te)
confusionMatrix(tree_tab_te)
cv_model <- cv.tree(tree_mod, FUN = prune.misclass)
plot(cv_model)
cv_model$dev # gives the deviance for each K (small is better)
best_size <- cv_model$size[which(cv_model$dev==min(cv_model$dev))] # which size is better?
best_size
# let's refit the tree model (the number of leafs will be no more than best.size)
cv_model_pruned <- prune.misclass(tree_mod, best=best_size)
summary(cv_model_pruned)
options(repr.plot.width=7, repr.plot.height=7, repr.plot.res = 200)
plot(cv_model_pruned)
text(cv_model_pruned)
pruned_tr_cv <- predict(cv_model_pruned, data_train, type="class") # give the predicted class
confusionMatrix(table(Prediction=pruned_tr_cv, Reference=data_train$diag1))
pruned_pred_cv <- predict(cv_model_pruned, data_test, type="class") # give the predicted class
confusionMatrix(table(Prediction=pruned_pred_cv, Reference=data_test$diag1))
################################################
pROC_obj <- roc(as.numeric(as.factor(data_test$diag1)),as.numeric(as.factor(tree_pred_te)),
smoothed = TRUE,
# arguments for ci
ci=TRUE, ci.alpha=0.95, stratified=FALSE,
# arguments for plot
plot=TRUE, auc.polygon=TRUE, max.auc.polygon=TRUE, grid=TRUE,
print.auc=TRUE, show.thres=TRUE)
sens.ci <- ci.se(pROC_obj)
plot(sens.ci, type="shape", col="lightblue")
###################################################################
################## Random Forest for Classification ###############
###################################################################
rf <- randomForest(factor(diag1) ~ ., mtry = 3, ntrees=500, importance = TRUE, data_train) ## suggested mtry=sqrt(p)
plot(rf)
summary(rf)
print(rf)
options(repr.plot.width=10, repr.plot.height=5, repr.plot.res = 200)
varImpPlot(rf, main="Variable Importance Accuracy and Gini coeff. from Random Forest")
importance(rf)
rf_pred_tr <- predict(rf, data_train[,-14], type='class')
rf_pred_te <- predict(rf, data_test[,-14], type='class')
rf_tab_tr <- table(Original=data_train$diag1, Prediction=rf_pred_tr)
rf_tab_te <- table(Original=data_test$diag1, Prediction=rf_pred_te)
confusionMatrix(rf_tab_tr)
confusionMatrix(rf_tab_te)
# importance(rf, type = 1)
# varImpPlot(rf, type = 1)
################################################
pROC_obj <- roc(as.numeric(as.factor(data_test$diag1)),as.numeric(as.factor(rf_pred_te)),
smoothed = TRUE,
# arguments for ci
ci=TRUE, ci.alpha=0.95, stratified=FALSE,
# arguments for plot
plot=TRUE, auc.polygon=TRUE, max.auc.polygon=TRUE, grid=TRUE,
print.auc=TRUE, show.thres=TRUE)
sens.ci <- ci.se(pROC_obj)
plot(sens.ci, type="shape", col="lightblue")
################## Iteration on mtry for RF #############
#########################################################
rf.c <- list()
yhat.rf <-list()
misclass_rf<-list()
for ( i in 1:13 ) {
set.seed(4444)
rf.c[[i]]<-randomForest(factor(diag1) ~ ., data = data_train, mtry = i, importance = TRUE)
yhat.rf[[i]]<-predict(rf.c[[i]], newdata = data_train[,-14])
misclass_rf[[i]]<- mean(yhat.rf[[i]] != data_test$diag1)
}
par(bg = "white" )
matplot(1:13, misclass_rf, xlab = 'No. of Variables (mtry)', ylab = 'Misclassification error', main = "Subset of predictors for Misclassification error")
lines(1:13, misclass_rf, type = "o")
## tree size estimation ##
rf_tree <- getTree(rf, k=2) # show the second tree
# print(rf_tree)
treesize(rf) # size of trees of the ensemble
mean(treesize(rf))
options(repr.plot.width=4, repr.plot.height=4, repr.plot.res = 200)
hist(treesize (rf), col = "red")
plotres(rf)
plotmo(rf, pmethod="partdep", all1=TRUE, all2=TRUE)
## end ##