RUGS-Meetup/baseFunctions_model.R at master · thiakx/RUGS-Meetup · 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
166
167
168
169
170
171
library(stringr)
library(caret)
library(data.table)

#load cleaned data
load("trainDataMod.Rdata")
load("testDataMod.Rdata")
load("trainData2.Rdata")
load("testData2.Rdata")

#only consider local moran and hex average if count of hex >30
trainDataMod$LM_avgView[trainDataMod$count<30]<-NA
trainDataMod$LM_avgVote[trainDataMod$count<30]<-NA
trainDataMod$LM_avgComment[trainDataMod$count<30]<-NA
trainDataMod<-na.omit(trainDataMod)
testDataMod<-na.omit(testDataMod)

#views, votes, comments each use thier own Local Moran
rowsNames<-names(trainDataMod)
drops <- c("hexNumber","id","num_votes","num_comments","num_views","count",
           "LM_count","LM_avgView","LM_avgComment","LM_avgVote",
           "avgComment","avgVote","avgView","year","month")
rowsNames<-rowsNames[!rowsNames %in% drops]

#use predicted views inside votes and comments as well, since logically, they are correlated
mod_viewsVar<-c(rowsNames,"LM_avgView")
mod_votesVar<-c(rowsNames,"num_views","LM_avgVote")
mod_commentsVar<-c(rowsNames,"num_views","LM_avgComment")


rfModel<-function(dataIn,dataTest,ntreeTotal,coreNumber){

  rf_views<- foreach(ntree=rep(ceiling(ntreeTotal/coreNumber),coreNumber), .combine=combine, .packages='randomForest',.inorder=FALSE) %dopar%
    randomForest(dataIn[,mod_viewsVar], dataIn$num_views, ntree=ntree)

  rf_votes<- foreach(ntree=rep(ceiling(ntreeTotal/coreNumber),coreNumber), .combine=combine, .packages='randomForest',.inorder=FALSE) %dopar%
    randomForest(dataIn[,mod_votesVar], dataIn$num_votes, ntree=ntree)

  rf_comments<- foreach(ntree=rep(ceiling(ntreeTotal/coreNumber),coreNumber), .combine=combine, .packages='randomForest',.inorder=FALSE) %dopar%
    randomForest(dataIn[,mod_commentsVar], dataIn$num_comments, ntree=ntree)

  rf_views_Test<-predict(rf_views, dataTest[,mod_viewsVar], type="response")
  dataTest$num_views<-rf_views_Test

  rf_votes_Test<-predict(rf_votes, dataTest[,mod_votesVar], type="response")

  rf_comments_Test<-predict(rf_comments, dataTest[,mod_commentsVar], type="response")
  dataTest$num_views<-NULL

  #only selected cols are chosen
  result<-cbind(dataTest[,c("id","city","tag_type","source")],rf_views_Test,rf_votes_Test,rf_comments_Test)
  names(result)<-c("id","city","tag_type","source","num_views","num_votes","num_comments")
  result<-result[,c("id","num_views","num_votes","num_comments","city","tag_type","source")]

  #undo the log for view
  result$num_views<-exp(result$num_views)-1

  #!< 0, votes cannot be less than 1
  result[result$num_votes<1,"num_votes"]<-1
  result[result$num_comments<0,"num_comments"]<-0
  result[result$num_views<0,"num_views"]<-0
  return (result)
}

glmModel<-function(dataIn,dataTest){

  glm_views<- glm(paste("num_views~",paste(mod_viewsVar, collapse = "+"),sep=""),data=dataIn)
  glm_votes<- glm(paste("num_votes~",paste(mod_votesVar, collapse = "+"),sep=""),data=dataIn)
  glm_comments<- glm(paste("num_comments~",paste(mod_commentsVar, collapse = "+"),sep=""),data=dataIn)

  glm_views_Test<-predict(glm_views, dataTest[,mod_viewsVar], type="response")
  dataTest$num_views<-glm_views_Test

  glm_votes_Test<-predict(glm_votes, dataTest[,mod_votesVar], type="response")

  glm_comments_Test<-predict(glm_comments, dataTest[,mod_commentsVar], type="response")
  dataTest$num_views<-NULL

  #only selected cols are chosen
  result<-cbind(dataTest[,c("id","city","tag_type","source")],glm_views_Test,glm_votes_Test,glm_comments_Test)
  names(result)<-c("id","city","tag_type","source","num_views","num_votes","num_comments")
  result<-result[,c("id","num_views","num_votes","num_comments","city","tag_type","source")]

  #undo the log
  result$num_views<-exp(result$num_views)-1

  #!< 0, votes cannot be less than 1
  result[result$num_votes<1,"num_votes"]<-1
  result[result$num_comments<0,"num_comments"]<-0
  result[result$num_views<0,"num_views"]<-0
  return (result)
}


gbmModel <-function(dataIn,dataTest){

#parameters to be tested
treeAmt     =10000
shinkAmt    =0.001
depthAmt    =5
minobsAmt   =25

#gbm submission
gbm_views<-gbm.fit(dataIn[,mod_viewsVar], dataIn$num_views,n.trees=treeAmt, shrinkage=shinkAmt, distribution="gaussian",
                   interaction.depth=depthAmt,n.minobsinnode = minobsAmt)

gbm_votes<-gbm.fit(dataIn[,mod_votesVar], dataIn$num_votes,n.trees=treeAmt, shrinkage=shinkAmt, distribution="poisson",
                interaction.depth=depthAmt,n.minobsinnode = minobsAmt)

gbm_comments<-gbm.fit(dataIn[,mod_commentsVar], dataIn$num_comments,n.trees=treeAmt, shrinkage=shinkAmt, distribution="poisson",
                   interaction.depth=depthAmt,n.minobsinnode = minobsAmt)

#oob best iter tend to under estimate, hence need to increase it by 20%
best.iter_views <- 1.2*gbm.perf(gbm_views,method="OOB")
best.iter_votes <- 1.2*gbm.perf(gbm_votes,method="OOB")
best.iter_comments <- 1.2*gbm.perf(gbm_comments,method="OOB")

gbm_views_Test <- predict(gbm_views,  dataTest[,mod_viewsVar],best.iter_views, type="response")
dataTest$num_views<-gbm_views_Test

gbm_votes_Test <- predict(gbm_votes,  dataTest[,mod_votesVar],best.iter_votes, type="response")

gbm_comments_Test <- predict(gbm_comments,  dataTest[,mod_commentsVar],best.iter_comments, type="response")
dataTest$num_views<-NULL

#only selected cols are chosen
result<-cbind(dataTest[,c("id","city","tag_type","source")],gbm_views_Test,gbm_votes_Test,gbm_comments_Test)
names(result)<-c("id","city","tag_type","source","num_views","num_votes","num_comments")
result<-result[,c("id","num_views","num_votes","num_comments","city","tag_type","source")]

#undo the log
result$num_views<-exp(result$num_views)-1

#!< 0, votes cannot be less than 1
result[result$num_votes<1,"num_votes"]<-1
result[result$num_comments<0,"num_comments"]<-0
result[result$num_views<0,"num_views"]<-0
return (result)

}


svmModel <-function(dataIn,dataTest){

  svm_views<- svm(as.formula(paste("num_views~",paste(mod_viewsVar, collapse = "+"),sep="")),data=dataIn)
  svm_votes<- svm(as.formula(paste("num_votes~",paste(mod_votesVar, collapse = "+"),sep="")),data=dataIn)
  svm_comments<- svm(as.formula(paste("num_comments~",paste(mod_commentsVar, collapse = "+"),sep="")),data=dataIn)

  svm_views_Test <- predict(svm_views,  dataTest[,mod_viewsVar],type="response")
  dataTest$num_views<-svm_views_Test

  svm_votes_Test <- predict(svm_votes,  dataTest[,mod_votesVar], type="response")

  svm_comments_Test <- predict(svm_comments,  dataTest[,mod_commentsVar], type="response")

  dataTest$num_views<-NULL

  #only selected cols are chosen
  result<-cbind(dataTest[,c("id","city","tag_type","source")],svm_views_Test,svm_votes_Test,svm_comments_Test)
  names(result)<-c("id","city","tag_type","source","num_views","num_votes","num_comments")
  result<-result[,c("id","num_views","num_votes","num_comments","city","tag_type","source")]

  #undo the log
  result$num_views<-exp(result$num_views)-1

  #!< 0, votes cannot be less than 1
  result[result$num_votes<1,"num_votes"]<-1
  result[result$num_comments<0,"num_comments"]<-0
  result[result$num_views<0,"num_views"]<-0
  return (result)
}