|
5
|
1 args <- commandArgs(T)
|
|
|
2
|
|
|
3 arg1 <- args[1]
|
|
|
4 arg2 <- args[2]
|
|
|
5 arg3 <- args[3]
|
|
|
6 arg4 <- args[4]
|
|
|
7 arg5 <- args[5]
|
|
|
8 arg6 <- args[6]
|
|
|
9 arg7 <- args[7]
|
|
|
10 arg8 <- args[8]
|
|
|
11 arg9 <- args[9]
|
|
|
12 library(caret)
|
|
|
13 load(arg1)
|
|
|
14
|
|
|
15 #RAWDATA <- dataX
|
|
|
16 #RAWDATA$outcome <- dataY
|
|
|
17
|
|
|
18
|
|
|
19 ###########################
|
|
|
20 Smpling <- arg9
|
|
|
21
|
|
|
22 if(Smpling=="downsampling")
|
|
|
23 {
|
|
|
24 dwnsmpl <- downSample(dataX,dataY)
|
|
|
25 RAWDATA <- dwnsmpl[,1:length(dwnsmpl)-1]
|
|
|
26 RAWDATA$outcome <- dwnsmpl[,length(dwnsmpl)]
|
|
|
27 dataX <- RAWDATA[,1:length(dwnsmpl)-1]
|
|
|
28 dataY <- RAWDATA[,"outcome"]
|
|
|
29 remove("dwnsmpl")
|
|
|
30 }else if(Smpling=="upsampling"){
|
|
|
31 upsmpl <- upSample(dataX,dataY)
|
|
|
32 RAWDATA <- upsmpl[,1:length(upsmpl)-1]
|
|
|
33 RAWDATA$outcome <- upsmpl[,length(upsmpl)]
|
|
|
34 dataX <- RAWDATA[,1:length(upsmpl)-1]
|
|
|
35 dataY <- RAWDATA[,"outcome"]
|
|
|
36 remove("upsmpl")
|
|
|
37 }else {
|
|
|
38 RAWDATA <- dataX
|
|
|
39 RAWDATA$outcome <- dataY
|
|
|
40 }
|
|
|
41
|
|
|
42
|
|
|
43
|
|
|
44
|
|
|
45 ##########################
|
|
|
46
|
|
|
47
|
|
|
48 rawData <- dataX
|
|
|
49 predictorNames <- names(rawData)
|
|
|
50
|
|
|
51 isNum <- apply(rawData[,predictorNames, drop = FALSE], 2, is.numeric)
|
|
|
52 if(any(!isNum)) stop("all predictors in rawData should be numeric")
|
|
|
53
|
|
|
54 colRate <- apply(rawData[, predictorNames, drop = FALSE],
|
|
|
55 2, function(x) mean(is.na(x)))
|
|
|
56 colExclude <- colRate > 0.1
|
|
|
57 if(any(colExclude)){
|
|
|
58 predictorNames <- predictorNames[-which(colExclude)]
|
|
|
59 rawData <- RAWDATA[, c(predictorNames,"outcome")]
|
|
|
60 } else {
|
|
|
61 rawData <- RAWDATA
|
|
|
62 }
|
|
|
63 rowRate <- apply(rawData[, predictorNames, drop = FALSE],
|
|
|
64 1, function(x) mean(is.na(x)))
|
|
|
65
|
|
|
66
|
|
|
67 rowExclude <- rowRate > 0
|
|
|
68 if(any(rowExclude)){
|
|
|
69 rawData <- rawData[!rowExclude, ]
|
|
|
70 ##hasMissing <- apply(rawData[, predictorNames, drop = FALSE],
|
|
|
71 ##1, function(x) mean(is.na(x)))
|
|
|
72
|
|
|
73 ############################################################################
|
|
|
74
|
|
|
75
|
|
|
76 ###############################################################################
|
|
|
77 } else {
|
|
|
78 rawData <- rawData[complete.cases(rawData),]
|
|
|
79
|
|
|
80 }
|
|
|
81
|
|
|
82 set.seed(2)
|
|
|
83
|
|
|
84 #print(dim(dataX))
|
|
|
85 #print(dim(rawData))
|
|
|
86 #print(length(dataY))
|
|
|
87
|
|
|
88 nzv <- nearZeroVar(rawData[,1:(length(rawData) - 1)])
|
|
|
89 if(length(nzv) > 0) {
|
|
|
90 #nzvVars <- names(rawData)[nzv]
|
|
|
91 rawData <- rawData[,-nzv]
|
|
|
92 #rawData$outcome <- dataY
|
|
|
93 }
|
|
|
94
|
|
|
95 predictorNames <- names(rawData)[names(rawData) != "outcome"]
|
|
|
96
|
|
|
97 dx <- rawData[,1:length(rawData)-1]
|
|
|
98 dy <- rawData[,length(rawData)]
|
|
|
99 corrThresh <- as.numeric(arg8)
|
|
|
100 highCorr <- findCorrelation(cor(dx, use = "pairwise.complete.obs"),corrThresh)
|
|
|
101 dx <- dx[, -highCorr]
|
|
|
102 subsets <- seq(1,length(dx),by=5)
|
|
|
103 normalization <- preProcess(dx)
|
|
|
104 dx <- predict(normalization, dx)
|
|
|
105 dx <- as.data.frame(dx)
|
|
|
106
|
|
|
107 if (arg4 == "lmFuncs"){
|
|
|
108 ctrl1 <- rfeControl(functions = lmFuncs,
|
|
|
109 method = arg5 ,
|
|
|
110 repeats = as.numeric(arg6),
|
|
|
111 number = as.numeric(arg7),
|
|
|
112 verbose = FALSE)
|
|
|
113 } else if(arg4 == "rfFuncs"){
|
|
|
114 ctrl1 <- rfeControl(functions = rfFuncs,
|
|
|
115 method = arg5 ,
|
|
|
116 repeats = as.numeric(arg6),
|
|
|
117 number = as.numeric(arg7),
|
|
|
118 verbose = FALSE)
|
|
|
119 }else if (arg4 == "treebagFuncs"){
|
|
|
120 ctrl1 <- rfeControl(functions = treebagFuncs,
|
|
|
121 method = arg5 ,
|
|
|
122 repeats = as.numeric(arg6),
|
|
|
123 number = as.numeric(arg7),
|
|
|
124 verbose = FALSE)
|
|
|
125 }else {
|
|
|
126
|
|
|
127 ctrl1 <- rfeControl(functions = nbFuncs,
|
|
|
128 method = arg5 ,
|
|
|
129 repeats = as.numeric(arg6),
|
|
|
130 number = as.numeric(arg7),
|
|
|
131 verbose = FALSE)
|
|
|
132 }
|
|
|
133
|
|
|
134
|
|
|
135
|
|
|
136
|
|
|
137 Profile <- rfe(dx, dy,sizes = subsets,rfeControl = ctrl1)
|
|
|
138
|
|
|
139 pred11 <- predictors(Profile)
|
|
|
140 save(Profile,file=arg2)
|
|
|
141 dataX <- rawData[,pred11]
|
|
|
142 dataY <- rawData$outcome
|
|
|
143
|
|
|
144 save(dataX,dataY,file=arg3)
|
|
|
145 rm(dataX)
|
|
|
146 rm(dataY)
|
|
|
147
|
