Repository navigation
Expand file tree
/
Copy pathweek12_assi6_sol1_Dhar.R
More file actions
437 lines (338 loc) · 14.5 KB
/
Copy pathweek12_assi6_sol1_Dhar.R
File metadata and controls
437 lines (338 loc) · 14.5 KB
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
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
# # -*- coding: utf-8 -*-
# """Week12 Assi6 Sol1.ipynb
#
# Automatically generated by Colaboratory.
#
# Original file is located at
# https://colab.research.google.com/drive/1BQQr3Sr0RTmqE_pa-jB5QKD6VN15PBgZ
# """
# ###########################################################################
## Week-12, Homework-6, Sol-1
## Sreya Dhar
## Created: Nov 24, 2020
## Edited: Dec 03, 2020
###########################################################################
# ## installing all the libaries 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('tree')
# install.packages('rpart')
# install.packages('rattle')
# install.packages("partykit")
# install.packages("randomForest")
# install.packages("party")
# install.packages("kernlab")
# install.packages("gbm")
# install.packages("plotmo")
# install.packages("pdp")
# """# New Section"""
## 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(caret)
library(MASS)
library(Hmisc)
library(hrbrthemes)
library(GGally)
library(glmnet)
library(pROC)
library(ISLR)
library(psych)
library(aod)
library(epiDisplay)
library(e1071)
library(ggthemes)
library(kableExtra)
library(broom)
library(knitr)
library(rpart)
library(rattle)
library(partykit)
library(randomForest)
library(tree)
library(party)
library(kernlab)
library(gbm)
library(plotmo)
library(tree)
library(pdp)
## uploading the data
data(OJ)
head(OJ)
head(OJ)
glimpse(OJ)
names(OJ)
status(OJ)
profiling_num(OJ)
describe(OJ)
summary(OJ)
options(repr.plot.width=8, repr.plot.height=8, repr.plot.res = 230)
pairs.panels(OJ[,2:18], main = "Pairs plot on OJ dataset, unclassed on Purchase", pch = 21, bg = c("blue", "green")[unclass(OJ$Purchase)], hist.col="red")
options(repr.plot.width=6, repr.plot.height=6, repr.plot.res = 200)
plot_num(OJ)
# OJ <- pima %>% mutate_if(is.character, as.factor)
OJ_n <- OJ %>% mutate_if(is.factor, as.numeric)
head(OJ_n)
# heatmap matrix
options(repr.plot.width=5, repr.plot.height=5, repr.plot.res = 230)
data_h <- as.data.frame(scale(OJ_n,center=TRUE,scale=TRUE))
heatmap.2(as.matrix(OJ_n), scale = "none", col = bluered(100), trace = "none", density.info = "none")
options(repr.plot.width=6, repr.plot.height=6, repr.plot.res = 200)
L <- cor(OJ_n)
corrplot(L, method = "circle", type = "lower")
########################################################################
##################### Classification Tree with rpart ###################
########################################################################
# Normalizing the predictor variable
# ## min-max scaling on vehicle dataset predictors prior to classification
normalized<-function(y) {
x<-y[!is.na(y)]
x<-(x - min(x)) / (max(x) - min(x))
y[!is.na(y)]<-x
return(y)
}
OJ[,2:18]<-apply(OJ_n[,c(2:18)],2,normalized)
head(OJ)
## splitting the dataset into train and test sets
set.seed(1234) ## seeding the sampling
data_split <- initial_split(OJ, prop = 0.75) ## spliting the data by library 'rsample'
data_train <- training(data_split)
data_test <- testing(data_split)
###################################################################
###################### Bagging ####################################
###################################################################
bag <- randomForest(factor(Purchase) ~ ., mtry = 17, ntrees=500, importance = TRUE, data_train) ## suggested mtry=sqrt(p)
bag_pred_tr <- predict(bag, data_train[,-1], type='class')
bag_pred_te <- predict(bag, data_test[,-1], type='class')
bag_tab_tr <- table(Original=data_train$Purchase, Prediction=bag_pred_tr)
bag_tab_te <- table(Original=data_test$Purchase, Prediction=bag_pred_te)
confusionMatrix(bag_tab_tr)
confusionMatrix(bag_tab_te)
importance(bag, type = 1)
varImpPlot(bag, type = 1)
##############################################################
#################### Random Forest ###########################
##############################################################
rf <- randomForest(factor(Purchase) ~., data=data_train, importance=TRUE, do.trace=100, ntree=500, mtry = 3)
print(rf)
rf$err.rate[500]
options(repr.plot.width=10, repr.plot.height=7, repr.plot.res = 200)
varImpPlot(rf, main="Variable Importance Accuracy and Gini coeff. from Random Forest")
importance(rf)
## Prediction on train set
rf_pred_tr <- predict(rf, data_train, type = "response")
rf_tab <- confusionMatrix(table(data_train$Purchase, rf_pred_tr))
rf_tab
## Prediction on test set
rf_predict <- predict(rf, data_test,type = "response")
rf_tab <- confusionMatrix(table(data_test$Purchase, rf_predict))
rf_tab
options(repr.plot.width=8, repr.plot.height=8, repr.plot.res = 230)
plotres(rf)
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")
######################################### Iteration on mtry for RF #########################################
rf.c <- list()
oob.err <- list()
yhat.rf <-list()
misclass_rf<-list()
for ( i in 1:17 ) {
set.seed(4444)
rf.c<-randomForest(Purchase ~ ., data = data_train, mtry = i, importance = TRUE, ntree = 500)
oob.err[i] = rf.c$err.rate[500] #Error of all Trees fitted
yhat.rf<-predict(rf.c, newdata = data_test[,-1])
misclass_rf[i]<- mean(yhat.rf != data_test$Purchase)
}
options(repr.plot.width=6, repr.plot.height=4, repr.plot.res = 200)
matplot(1:17, misclass_rf, xlab = 'No. of Variables (mtry)', ylab = 'Misclassification error', main = "Subset size Vs. Misclassification error")
lines(1:17, misclass_rf, type = "o")
options(repr.plot.width=7, repr.plot.height=6, repr.plot.res = 200)
matplot(1:17 , cbind(oob.err, misclass_rf), pch=19 , col=c("red","blue"), type=c("o","o"), ylab="Misclassification Error rate", xlab="Number of Predictors Considered at each Split, mtry", lty=1, lwd=2)
legend("topright", legend=c("Out of Bag Error","Test Error"), pch=19, col=c("red","blue"))
misclass_rf ## min for mtry=4,7 on test set
################ Tuning on RF model ###############
library("e1071") # to 'tune' the rf model
tuned_rf <- tune(randomForest, train.x = Purchase ~ ., data = data_train, validation.x = data_test, mtry=4, ntree=500)
best_rf <- tuned_rf$best.model
best_rf_pred <- predict(best_rf, data_test, type='response')
tab_best_rf <- table(Reference=data_test$Purchase, Prediction=best_rf_pred)
confusionMatrix(tab_best_rf)
tuned_rf
best_rf
options(repr.plot.width=10, repr.plot.height=5, repr.plot.res = 200)
par(mfrow=c(1,2))
plot(rf)
plot(best_rf)
# computing overall error:
err_rf <- 1 - sum(diag(as.matrix(tab_best_rf))) / sum(tab_best_rf)
err_rf
options(repr.plot.width=8, repr.plot.height=8, repr.plot.res = 230)
plotmo(rf, pmethod="partdep", all1=TRUE )
data_train_n <- data_train %>% mutate_if(is.factor, as.numeric)-1
data_test_n <- data_test %>% mutate_if(is.factor, as.numeric)-1
###############################################
# Boosting ##
###############################################
mod_boost = gbm(Purchase ~.,
data = data_train_n,
distribution = "adaboost",
cv.folds = 10,
shrinkage = .1,
n.minobsinnode = 10,
n.trees = 200)
print(mod_boost)
options(repr.plot.width=8, repr.plot.height=8, repr.plot.res = 230)
plotres(mod_boost)
mod_boost1 = gbm(Purchase ~.,
data = data_train,
distribution = "multinomial",
cv.folds = 10,
shrinkage = .1,
n.minobsinnode = 10,
n.trees = 200)
print(mod_boost1)
### prediction on train set from boosting ###
pred_boost_tr = predict.gbm(object = mod_boost1,
newdata = data_train,
n.trees = 200,
type = "response")
labels_tr = colnames(pred_boost_tr)[apply(pred_boost_tr, 1, which.max)]
result_tr = data.frame(data_train$Purchase, labels_tr)
confusionMatrix(data_train$Purchase, as.factor(labels_tr))
### prediction on test set from boosting ###
pred_boost_te = predict.gbm(object = mod_boost1,
newdata = data_test,
n.trees = 200,
type = "response")
labels = colnames(pred_boost_te)[apply(pred_boost_te, 1, which.max)]
result = data.frame(data_test$Purchase, labels)
# print(result)
confusionMatrix(data_test$Purchase, as.factor(labels))
#### Shrinkage factor iteration for boosting ###
shrink <- c(.1, .4, .6, .8)
max_iter <- 5000
store_error <- c()
for (i in 1:length(shrink)){
boost.fit <- gbm(Purchase~., data = data_train_n, n.trees = max_iter, shrinkage = shrink[i], interaction.depth = 3, distribution = "adaboost")
temp <- c()
for (j in 1:max_iter){
y_hat <- predict(boost.fit, newdat = data_test_n, n.trees = j, type = "response")
misclass_boost <- mean(y_hat != data_test_n$Purchase)
temp <- c(temp, misclass_boost)
}
store_error <- cbind(store_error, temp) # max_iter x length(shrink)
}
colnames(store_error) <- paste("shrinkage", shrink, sep = ":")
options(repr.plot.width=7, repr.plot.height=7, repr.plot.res = 200)
plot(store_error[,1], type = "l", main = "Error Profiles", ylab = "error", xlab = "boosting iterations", ylim=c(0.955,1))
lines(store_error[,2], col = "red")
lines(store_error[,3], col = "blue")
lines(store_error[,4], col = "green")
legend("bottomleft", legend=c("shrinkage = 0.1","shrinkage = 0.4", "shrinkage = 0.6", "shrinkage = 0.8"), pch=19, col=c('black',"red","blue", "green"))
options(repr.plot.width=12, repr.plot.height=10, repr.plot.res = 230)
plotmo(mod_boost, pmethod="partdep", all1=TRUE)
###################################################################
############ Knn prediction on the dataset ########################
###################################################################
accuracy = function(actual, predicted) { ## defining accuracy function
mean(actual == predicted)}
error = function(actual, predicted) { ## defining error function
mean(actual != predicted)}
iter_k = c(1,3,5,7,9,13,15) ## giving k-values
accu_pca_te = rep(x = 0, times = length(iter_k))
accu_pca_tr = rep(x = 0, times = length(iter_k))
error_pca_te = rep(x = 0, times = length(iter_k))
error_pca_tr = rep(x = 0, times = length(iter_k))
for(i in seq_along(iter_k)) {
pred_pca_tr = knn( train = data_train[,2:18],
test = data_train[,2:18],
cl = data_train$Purchase,
k = iter_k[i])
accu_pca_tr[i] = accuracy(pred_pca_tr,data_train$Purchase) ## accuracy from knn on train set
error_pca_tr[i] = error( pred_pca_tr, data_train$Purchase) ## error from knn on train set
} ## error from knn train set
for(i in seq_along(iter_k)) {
pred_pca_te = knn( train = data_train[,2:18],
test = data_test[,2:18],
cl = data_train$Purchase,
k = iter_k[i])
accu_pca_te[i] = accuracy(pred_pca_te, data_test$Purchase) ## accuracy from knn on test set
error_pca_te[i] = error(pred_pca_te, data_test$Purchase) ## error from knn on test set
} ## error from knn test set
error_pca_train <- mean(data_train$Purchase != pred_pca_tr)
error_pca_test <- mean(data_test$Purchase != pred_pca_te)
print(paste('Accuracy of train set from KNN',(1-error_pca_train)*100,'%'))
print(paste('Accuracy of test set from KNN',(1-error_pca_test)*100,'%'))
c(error_pca_te, error_pca_tr)
####################### Comparing classification accuracy and errror in knn from PCA components #########################
options(repr.plot.width=8, repr.plot.height=8, repr.plot.res = 200)
par(mfrow=c(2,2))
# plot accuracy vs choice of k on Training set
plot(iter_k, accu_pca_tr*100, type = "b",col = "blue", cex = 1, pch = 20, lwd = 2,
xlab = "k, number of neighbors", ylim= c(70,100),
ylab = "classification accuracy, %", main = "Accuracy of Training set")
abline(v = which.max(accu_pca_tr),y = max(accu_pca_tr)*100, type = "l", col = "red", lwd = 2, lty = 2)
abline(x = which.max(accu_pca_tr),h = max(accu_pca_tr)*100, type = "l", col = "black", lty = 2)
# plot accuracy vs choice of k on Test set
plot(iter_k, accu_pca_te*100, type = "b", col = "blue", cex = 1, pch = 20, lwd = 2,
ylim= c(70,100),
xlab = "k, number of neighbors", ylab = "classification accuracy, %",
main = "Accuracy of Test set")
abline(v = 9,y = max(accu_pca_te)*100, type = "l", col = "red", lwd = 2, lty = 2)
abline(x = which.max(accu_pca_te),h = max(accu_pca_te)*100, type = "l", col = "black", lty = 2)
# plot accuracy vs choice of k on Training set
plot(iter_k, error_pca_tr*100, type = "b",col = "blue", cex = 1, pch = 20, lwd = 2,
xlab = "k, number of neighbors", ylim=c(0,30),
ylab = "classification error, %", main = "classification error of Training set")
abline(v = which.min(error_pca_tr),y = min(error_pca_tr)*100, type = "l", col = "red", lwd = 2, lty = 2)
abline(x = which.min(error_pca_tr),h = min(error_pca_tr)*100, type = "l", col = "black", lty = 2)
# # plot accuracy vs choice of k on Test set
plot(iter_k, error_pca_te*100, type = "b", col = "blue", cex = 1, pch = 20, lwd = 2,
ylim=c(0,30),
xlab = "k, number of neighbors", ylab = "classification error, %",
main = "classification error of Test set")
abline(v = 9,y = min(error_pca_te)*100, type = "l", col = "red", lwd = 2, lty = 2)
abline(x = which.min(error_pca_te),h = min(error_pca_te)*100, type = "l", col = "black", lty = 2)
### end ###