-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathna_grnn_cli.R
172 lines (147 loc) · 6.29 KB
/
na_grnn_cli.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
# 22.3.17 IVA
# http://stackoverflow.com/questions/18695335/replacing-all-nas-with-smoothing-spline
# install.packages('grnn', dependencies=TRUE, repos='http://cran.rstudio.com/')
# install.packages('grt', dependencies=TRUE, repos='http://cran.rstudio.com/')
# install.packages('neuralnet', dependencies=TRUE, repos='http://cran.rstudio.com/')
# install.packages('caret', dependencies=TRUE, repos='http://cran.rstudio.com/')
require(zoo)
require(grnn)
require(grt)
require(foreach)
is.leapyear=function(year){
#http://en.wikipedia.org/wiki/Leap_year
#https://www.r-bloggers.com/leap-years/
#https://github.com/hadley/lubridate/blob/master/R/leap-years.r
return(((year %% 4 == 0) & (year %% 100 != 0)) | (year %% 400 == 0))
}
setZeroNegativeNumber <- function(n){
if(n < 0) return(0)
else return(n)
}
createCliDF <- function(fileName){
cli_dataset <-read.table(fileName, header=FALSE, sep="")
names(cli_dataset) <- c("day", "month", "year", "prec", "temp");
cli_dataset[cli_dataset$prec == -9999, 4] <- NA ; prec_vector <- cli_dataset[, 4]
cli_dataset[cli_dataset$temp == -9999, 5] <- NA ; temp_vector <- cli_dataset[, 5]
days <- c(1:dim(cli_dataset)[1])
return(data.frame(days=days, prec=prec_vector, temp=temp_vector))
}
# Replacement numbers-9999 to missing value NA
replacementNumbersMinus9999.NA <- function(cli_dataset){
for(i in 1:nrow(cli_dataset)){
if(!is.na(cli_dataset[i,2])){
if(cli_dataset[i,2]==-9999) cli_dataset[i,2]<-NA
}
if(!is.na(cli_dataset[i,3])) {
if(cli_dataset[i,3]==-9999) cli_dataset[i,3]<-NA
}
}
return(cli_dataset)
}
plotTempFilled <- function(naDF, filledDF, fileName){
plot(filledDF[,3], main=fileName, ylab="Temp(c)*10", xlab="Days", col="red", pch=19);
points(naDF[,3], col="blue", pch=19)
grid (10,10, lty = 6, col = "cornsilk2")
points(filledDF[,3], col="red", pch=19);
points(naDF[,3], col="blue", pch=19)
# Add a legend
legend("topleft", legend=c("Real CLI", "Guess CLI"), col=c("red", "blue"), pch=c(19,19), cex=0.8)
}
plotPrecFilled <- function(naDF, filledDF, fileName){
plot(filledDF[,2], main=fileName, ylab="Prec(mm)*10", xlab="Days", col="red", pch=19);
points(naDF[,2], col="blue", pch=19)
grid (10,10, lty = 6, col = "cornsilk2")
points(filledDF[,2], col="red", pch=19);
points(naDF[,2], col="blue", pch=19)
# Add a legend
legend("topleft", legend=c("Real CLI", "Guess CLI"), col=c("red", "blue"), pch=c(19,19), cex=0.8)
}
na.grnn <- function(vec.cli, setTrain, setSigma, searchSigma, trace){
# PRE-PROCESSING DATA
vec.na<-vec.cli; vec.na.scale <- grt::scale(vec.na);
vec.na.scale.min<-min(vec.na.scale, na.rm = T); vec.na.scale.max<-max(vec.na.scale, na.rm = T);
vec.na.index <- which(is.na(vec.na)) #
vec.na.scale.index <- which(is.na(vec.na.scale)) #
vec.na.scale.na.omit <- na.omit(vec.na.scale);
days <- 1:length(vec.na); days.scale <- grt::scale(days) # äëÿ ïðîãíîçà
days.scale.na.omit <- days.scale[-vec.na.scale.index]; # base::scale(days)
XY <- data.frame(days.scale.na.omit, vec.na.scale.na.omit)
if(searchSigma == TRUE){
## SPLIT DATA SAMPLES
#seed.init<-2017; set.seed(seed.init)
num.test<-setTrain;
rows <- sample(1:nrow(XY), nrow(XY) - num.test)
trainXY <- XY[rows, ]; testXY <- XY[-rows, ]
s.seq <- seq(0.01, 0.99, 0.05)
cv <- foreach(s = s.seq) %do% {
L <- grnn::learn(trainXY, variable.column = ncol(trainXY))
grnn <- grnn::smooth(L, sigma = s)
guessY <- testXY[, 2]
for(i in 1:length(testXY[, 1])){
guessY[i] <- grnn::guess(grnn, testXY[i, 1])
}
test.sse <- round(sum((guessY-testXY[, 2]))^2, 6)
if(trace == TRUE){
cat("Sigma= ", round(s,3), "SSE= ", test.sse, "\n")
}
data.frame(s, sse = test.sse)
# http://r-train.ru/r-%D0%BA%D0%B2%D0%B0%D0%B4%D1%80%D0%B0%D1%82-rse-%D0%B8-rmse/
}
idx<-1; val<-cv[[1]][1,2]
for(i in 2: length(s.seq)){
if(cv[[i]][1,2] < val){
idx <- i
val <- cv[[i]][1,2]
}
}
idx; val;
## predicting missing values with the calculated optimum Sigma GRNN
s<-cv[[idx]][1,1]
} else s <- setSigma
L <- grnn::learn(XY, variable.column = ncol(XY))
grnn <- grnn::smooth(L, sigma = s)
for(i in vec.na.index){
#todo ïîñòàâèòü ïðîâåðêó âûõîäà çà min max
vec.na.scale[i] <- grnn::guess(grnn, days.scale[i, 1])
#cat(i, vec.na.scale[i],"\n")
}
vec.na.unscale <- grt::unscale(vec.na.scale)
#plot(vec.na.unscale, col="red"); points(vec.na, col="blue");
}
na.grnn.cli <- function(cli_dataset, setTrain, setSigma, searchSigma, trace){
cli_dataset<-replacementNumbersMinus9999.NA(cli_dataset)
prec_vector <- cli_dataset[, 2]
temp_vector <- cli_dataset[, 3]
#days <- length(cli_dataset[,1])
days <- c(1:dim(cli_dataset)[1])
if(sum(is.na(prec_vector)) != 0){
prec_filled <- sapply(as.integer(na.grnn(prec_vector, setTrain, setSigma, searchSigma, trace)), setZeroNegativeNumber)
} else prec_filled <- prec_vector
if(sum(is.na(temp_vector)) != 0){
#temp_filled <- as.integer(na.spline(temp_vector))
temp_filled <- as.integer(na.grnn(temp_vector, setTrain, setSigma, searchSigma, trace))
} else temp_filled <- temp_vector
new_cli <- data.frame(day=cli_dataset[,1], prec=prec_filled, temp=temp_filled)
return(new_cli)
}
na.grnn.cli.test <- function(fileName, setTrain, setSigma, searchSigma, trace){
filled_path <- "C:/Users/IVA/Dropbox/Apps/na_grnn"
#filled_path <- "/home/larisa/Dropbox/Apps/na_grnn"
setwd(filled_path); getwd(); file_name <- fileName
cli_dataset <- createCliDF(file_name)
summary(cli_dataset)
#system.time(na.na<-na.grnn(cli_dataset[, 3]))
#user system elapsed Atom home
#23.52 2.03 32.03
#ïîëüçîâàòåëü ñèñòåìà ïðîøëî
#3.43 0.00 3.81
cli_dataset_filled <- na.grnn.cli(cli_dataset, setTrain, setSigma, searchSigma, trace)
plotTempFilled(cli_dataset, cli_dataset_filled, file_name)
#plotPrecFilled(cli_dataset, cli_dataset_filled, file_name)
}
### TEST SCRIPT
if(FALSE){
# https://gist.githubusercontent.com/anonymous/75b0aa2a79091549247e6596f989712e/raw/def3b6967dc8c1a527a0472012cf57a44b1b75ab/1967.cli
na.grnn.cli.test("1967na.cli", 67, 0.01, TRUE, TRUE)
na.grnn.cli.test("1967na.cli", 67, 0.01, TRUE, FALSE)
}