1.
## 13.3 보스톤 집값 예측
# 데이터 확인
install.packages("neuralnet")
install.packages("MASS")
library(neuralnet)
library(MASS)
{
head(Boston)
tail(Boston)
# 분석 대상 데이터 할당
# crim, zn, chas, nox, rad, mediv
data <- Boston[, c(1,2,4,5,9,14)]
head(data)
}
{
# 결측치 확인
na <- apply(data, 2, is.na)
na
apply(na, 2, sum)
# 데이터 정규화
maxs <- apply(data, 2, max)
maxs
mins <- apply(data, 2, min)
mins
data.scaled <- scale(data, center = mins, scale = maxs - mins)
data.scaled
}
{
# 학습용과 테스트용 데이터 분리
n <- nrow(data.scaled)
n
set.seed(1234)
index <- sample(1:n, round(0.8*n))
index
train <- as.data.frame(data.scaled[index,])
head(train)
test <- as.data.frame(data.scaled[-index,])
head(test)
}
{
# 학습용 입출력 변수 할당
names.col <- colnames(train)
names.col
var.dependent <- names.col[6]
var.dependent
var.independent <- names.col[-6]
var.independent
f.var.independent <- paste(var.independent, collapse = " + ")
f.var.independent
f <- paste(var.dependent, "~", f.var.independent)
f
}
{
# 학습 및 모형 출력
model <- neuralnet(f,
data = train,
hidden = c(5, 3, 3, 2),
linear.output = T)
plot(model)
}
{
# 테스트
predicted <- compute(model, test)
predicted$net.result
predicted.real <- predicted$net.result * (maxs[6] - mins[6]) + mins[6]
predicted.real
test.real <- test$medv * (maxs[6] - mins[6]) + mins[6]
test.real
}
{
# 실제값 대비 예측값 비교(분포)
plot(test.real, predicted.real,
xlim=c(0, 50), ylim=c(0, 50),
main="실제값 대비 예측값 분포",
xlab="실제값", ylab="예측값",
col="red",
pch=18, cex=0.7)
abline(0, 1, col="blue", lty=2)
MAPE.model <- sum(abs(test.real - predicted.real)/test.real * 100) / length(test.real)
MAPE.model
}실행결과
# 데이터 확인
> install.packages("neuralnet")
WARNING: Rtools is required to build R packages but is not currently installed. Please download and install the appropriate version of Rtools before proceeding:
https://cran.rstudio.com/bin/windows/Rtools/
‘C:/Users/PC44069/Documents/R/win-library/4.1’의 위치에 패키지(들)을 설치합니다.
(왜냐하면 ‘lib’가 지정되지 않았기 때문입니다)
‘Deriv’(들)을 또한 설치합니다.
trying URL 'https://cran.rstudio.com/bin/windows/contrib/4.1/Deriv_4.1.3.zip'
Content type 'application/zip' length 148835 bytes (145 KB)
downloaded 145 KB
trying URL 'https://cran.rstudio.com/bin/windows/contrib/4.1/neuralnet_1.44.2.zip'
Content type 'application/zip' length 123926 bytes (121 KB)
downloaded 121 KB
package ‘Deriv’ successfully unpacked and MD5 sums checked
package ‘neuralnet’ successfully unpacked and MD5 sums checked
The downloaded binary packages are in
C:\Users\Public\Documents\ESTsoft\CreatorTemp\Rtmp6rxTuC\downloaded_packages
> install.packages("MASS")
WARNING: Rtools is required to build R packages but is not currently installed. Please download and install the appropriate version of Rtools before proceeding:
https://cran.rstudio.com/bin/windows/Rtools/
‘C:/Users/PC44069/Documents/R/win-library/4.1’의 위치에 패키지(들)을 설치합니다.
(왜냐하면 ‘lib’가 지정되지 않았기 때문입니다)
trying URL 'https://cran.rstudio.com/bin/windows/contrib/4.1/MASS_7.3-57.zip'
Content type 'application/zip' length 1189664 bytes (1.1 MB)
downloaded 1.1 MB
package ‘MASS’ successfully unpacked and MD5 sums checked
The downloaded binary packages are in
C:\Users\Public\Documents\ESTsoft\CreatorTemp\Rtmp6rxTuC\downloaded_packages
> library(neuralnet)
> library(MASS)
> {
+ head(Boston)
+ tail(Boston)
+
+ # 분석 대상 데이터 할당
+ # crim, zn, chas, nox, rad, mediv
+ data <- Boston[, c(1,2,4,5,9,14)]
+ head(data)
+ }
crim zn chas nox rad medv
1 0.00632 18 0 0.538 1 24.0
2 0.02731 0 0 0.469 2 21.6
3 0.02729 0 0 0.469 2 34.7
4 0.03237 0 0 0.458 3 33.4
5 0.06905 0 0 0.458 3 36.2
6 0.02985 0 0 0.458 3 28.7
> {
+ # 결측치 확인
+ na <- apply(data, 2, is.na)
+ na
+ apply(na, 2, sum)
+
+ # 데이터 정규화
+ maxs <- apply(data, 2, max)
+ maxs
+ mins <- apply(data, 2, min)
+ mins
+ data.scaled <- scale(data, center = mins, scale = maxs - mins)
+ data.scaled
+ }
crim zn chas nox rad medv
1 0.000000e+00 0.180 0 0.314814815 0.00000000 0.42222222
2 2.359225e-04 0.000 0 0.172839506 0.04347826 0.36888889
3 2.356977e-04 0.000 0 0.172839506 0.04347826 0.66000000
4 2.927957e-04 0.000 0 0.150205761 0.08695652 0.63111111
5 7.050701e-04 0.000 0 0.150205761 0.08695652 0.69333333
6 2.644715e-04 0.000 0 0.150205761 0.08695652 0.52666667
7 9.213230e-04 0.125 0 0.286008230 0.17391304 0.39777778
8 1.553672e-03 0.125 0 0.286008230 0.17391304 0.49111111
9 2.303251e-03 0.125 0 0.286008230 0.17391304 0.25555556
10 1.840173e-03 0.125 0 0.286008230 0.17391304 0.30888889
11 2.456674e-03 0.125 0 0.286008230 0.17391304 0.22222222
12 1.249299e-03 0.125 0 0.286008230 0.17391304 0.30888889
13 9.830293e-04 0.125 0 0.286008230 0.17391304 0.37111111
14 7.007315e-03 0.000 0 0.314814815 0.13043478 0.34222222
15 7.099481e-03 0.000 0 0.314814815 0.13043478 0.29333333
16 6.980677e-03 0.000 0 0.314814815 0.13043478 0.33111111
17 1.177488e-02 0.000 0 0.314814815 0.13043478 0.40222222
18 8.743184e-03 0.000 0 0.314814815 0.13043478 0.27777778
19 8.951232e-03 0.000 0 0.314814815 0.13043478 0.33777778
20 8.086782e-03 0.000 0 0.314814815 0.13043478 0.29333333
21 1.399878e-02 0.000 0 0.314814815 0.13043478 0.19111111
22 9.505689e-03 0.000 0 0.314814815 0.13043478 0.32444444
23 1.378163e-02 0.000 0 0.314814815 0.13043478 0.22666667
24 1.103868e-02 0.000 0 0.314814815 0.13043478 0.21111111
25 8.361706e-03 0.000 0 0.314814815 0.13043478 0.23555556
26 9.376432e-03 0.000 0 0.314814815 0.13043478 0.19777778
27 7.481071e-03 0.000 0 0.314814815 0.13043478 0.25777778
28 1.067159e-02 0.000 0 0.314814815 0.13043478 0.21777778
29 8.617186e-03 0.000 0 0.314814815 0.13043478 0.29777778
30 1.119626e-02 0.000 0 0.314814815 0.13043478 0.35555556
31 1.263900e-02 0.000 0 0.314814815 0.13043478 0.17111111
32 1.515569e-02 0.000 0 0.314814815 0.13043478 0.21111111
33 1.552964e-02 0.000 0 0.314814815 0.13043478 0.18222222
34 1.287402e-02 0.000 0 0.314814815 0.13043478 0.18000000
35 1.805667e-02 0.000 0 0.314814815 0.13043478 0.18888889
36 6.502201e-04 0.000 0 0.234567901 0.17391304 0.30888889
37 1.024167e-03 0.000 0 0.234567901 0.17391304 0.33333333
38 8.297190e-04 0.000 0 0.234567901 0.17391304 0.35555556
39 1.896485e-03 0.000 0 0.234567901 0.17391304 0.43777778
40 2.395193e-04 0.750 0 0.088477366 0.08695652 0.57333333
41 3.065082e-04 0.750 0 0.088477366 0.08695652 0.66444444
42 1.361360e-03 0.000 0 0.129629630 0.08695652 0.48000000
43 1.519391e-03 0.000 0 0.129629630 0.08695652 0.45111111
44 1.720133e-03 0.000 0 0.129629630 0.08695652 0.43777778
45 1.307971e-03 0.000 0 0.129629630 0.08695652 0.36000000
46 1.855684e-03 0.000 0 0.129629630 0.08695652 0.31777778
47 2.046086e-03 0.000 0 0.129629630 0.08695652 0.33333333
48 2.505904e-03 0.000 0 0.129629630 0.08695652 0.25777778
49 2.782402e-03 0.000 0 0.129629630 0.08695652 0.20888889
50 2.399127e-03 0.000 0 0.129629630 0.08695652 0.32000000
51 9.262685e-04 0.210 0 0.111111111 0.13043478 0.32666667
52 4.164331e-04 0.210 0 0.111111111 0.13043478 0.34444444
53 5.314158e-04 0.210 0 0.111111111 0.13043478 0.44444444
54 4.888171e-04 0.210 0 0.111111111 0.13043478 0.40888889
55 8.182544e-05 0.750 0 0.051440329 0.08695652 0.30888889
56 7.631796e-05 0.900 0 0.037037037 0.17391304 0.67555556
57 1.599418e-04 0.850 0 0.051440329 0.04347826 0.43777778
58 8.991807e-05 1.000 0 0.053497942 0.17391304 0.59111111
59 1.664945e-03 0.250 0 0.139917695 0.30434783 0.40666667
60 1.089807e-03 0.250 0 0.139917695 0.30434783 0.32444444
61 1.607286e-03 0.250 0 0.139917695 0.30434783 0.30444444
62 1.858944e-03 0.250 0 0.139917695 0.30434783 0.24444444
63 1.168373e-03 0.250 0 0.139917695 0.30434783 0.38222222
64 1.350794e-03 0.250 0 0.139917695 0.30434783 0.44444444
65 1.482524e-04 0.175 0 0.063991770 0.08695652 0.62222222
66 3.317977e-04 0.800 0 0.026748971 0.13043478 0.41111111
67 4.211538e-04 0.800 0 0.026748971 0.13043478 0.32000000
68 5.796344e-04 0.125 0 0.049382716 0.13043478 0.37777778
69 1.452402e-03 0.125 0 0.049382716 0.13043478 0.27555556
70 1.369452e-03 0.125 0 0.049382716 0.13043478 0.35333333
71 9.209858e-04 0.000 0 0.057613169 0.13043478 0.42666667
72 1.713389e-03 0.000 0 0.057613169 0.13043478 0.37111111
73 9.589762e-04 0.000 0 0.057613169 0.13043478 0.39555556
74 2.125101e-03 0.000 0 0.057613169 0.13043478 0.40888889
75 8.164561e-04 0.000 0 0.106995885 0.17391304 0.42444444
76 9.980906e-04 0.000 0 0.106995885 0.17391304 0.36444444
77 1.070137e-03 0.000 0 0.106995885 0.17391304 0.33333333
78 9.076105e-04 0.000 0 0.106995885 0.17391304 0.35111111
79 5.635615e-04 0.000 0 0.106995885 0.17391304 0.36000000
80 8.716433e-04 0.000 0 0.106995885 0.17391304 0.34000000
81 3.912560e-04 0.250 0 0.084362140 0.13043478 0.51111111
82 4.304828e-04 0.250 0 0.084362140 0.13043478 0.42000000
83 3.402275e-04 0.250 0 0.084362140 0.13043478 0.44000000
84 3.280886e-04 0.250 0 0.084362140 0.13043478 0.39777778
85 4.975841e-04 0.000 0 0.131687243 0.08695652 0.42000000
86 5.735649e-04 0.000 0 0.131687243 0.08695652 0.48000000
87 5.120834e-04 0.000 0 0.131687243 0.08695652 0.38888889
88 7.327199e-04 0.000 0 0.131687243 0.08695652 0.38222222
89 5.651351e-04 0.000 0 0.213991770 0.04347826 0.41333333
90 5.248967e-04 0.000 0 0.213991770 0.04347826 0.52666667
91 4.554350e-04 0.000 0 0.213991770 0.04347826 0.39111111
92 3.709120e-04 0.000 0 0.213991770 0.04347826 0.37777778
93 4.013718e-04 0.280 0 0.162551440 0.13043478 0.39777778
94 2.521078e-04 0.280 0 0.162551440 0.13043478 0.44444444
95 4.116000e-04 0.280 0 0.162551440 0.13043478 0.34666667
96 1.300665e-03 0.000 0 0.123456790 0.04347826 0.52000000
97 1.221987e-03 0.000 0 0.123456790 0.04347826 0.36444444
98 1.287065e-03 0.000 0 0.123456790 0.04347826 0.74888889
99 8.491638e-04 0.000 0 0.123456790 0.04347826 0.86222222
100 7.000122e-04 0.000 0 0.123456790 0.04347826 0.62666667
101 1.599867e-03 0.000 0 0.277777778 0.17391304 0.50000000
102 1.213894e-03 0.000 0 0.277777778 0.17391304 0.47777778
103 2.500172e-03 0.000 0 0.277777778 0.17391304 0.30222222
104 2.307410e-03 0.000 0 0.277777778 0.17391304 0.31777778
105 1.498035e-03 0.000 0 0.277777778 0.17391304 0.33555556
106 1.419582e-03 0.000 0 0.277777778 0.17391304 0.32222222
107 1.853211e-03 0.000 0 0.277777778 0.17391304 0.32222222
108 1.403284e-03 0.000 0 0.277777778 0.17391304 0.34222222
109 1.367879e-03 0.000 0 0.277777778 0.17391304 0.32888889
110 2.892102e-03 0.000 0 0.277777778 0.17391304 0.32000000
111 1.142072e-03 0.000 0 0.277777778 0.17391304 0.37111111
112 1.062382e-03 0.000 0 0.333333333 0.21739130 0.39555556
113 1.314715e-03 0.000 0 0.333333333 0.21739130 0.30666667
114 2.425540e-03 0.000 0 0.333333333 0.21739130 0.30444444
115 1.528495e-03 0.000 0 0.333333333 0.21739130 0.30000000
116 1.854785e-03 0.000 0 0.333333333 0.21739130 0.29555556
117 1.407892e-03 0.000 0 0.333333333 0.21739130 0.36000000
118 1.625944e-03 0.000 0 0.333333333 0.21739130 0.31555556
119 1.396652e-03 0.000 0 0.333333333 0.21739130 0.34222222
120 1.556032e-03 0.000 0 0.333333333 0.21739130 0.31777778
121 7.043957e-04 0.000 0 0.403292181 0.04347826 0.37777778
122 7.342934e-04 0.000 0 0.403292181 0.04347826 0.34000000
123 9.741499e-04 0.000 0 0.403292181 0.04347826 0.34444444
124 1.619200e-03 0.000 0 0.403292181 0.04347826 0.27333333
125 1.035969e-03 0.000 0 0.403292181 0.04347826 0.30666667
126 1.828709e-03 0.000 0 0.403292181 0.04347826 0.36444444
127 4.282685e-03 0.000 0 0.403292181 0.04347826 0.23777778
128 2.841748e-03 0.000 0 0.491769547 0.13043478 0.24888889
129 3.586719e-03 0.000 0 0.491769547 0.13043478 0.28888889
130 9.834002e-03 0.000 0 0.491769547 0.13043478 0.20666667
131 3.751157e-03 0.000 0 0.491769547 0.13043478 0.31555556
132 1.333732e-02 0.000 0 0.491769547 0.13043478 0.32444444
133 6.560984e-03 0.000 0 0.491769547 0.13043478 0.40000000
134 3.636062e-03 0.000 0 0.491769547 0.13043478 0.29777778
135 1.090088e-02 0.000 0 0.491769547 0.13043478 0.23555556
136 6.198277e-03 0.000 0 0.491769547 0.13043478 0.29111111
137 3.555361e-03 0.000 0 0.491769547 0.13043478 0.27555556
138 3.889069e-03 0.000 0 0.491769547 0.13043478 0.26888889
139 2.736656e-03 0.000 0 0.491769547 0.13043478 0.18444444
140 6.049238e-03 0.000 0 0.491769547 0.13043478 0.28444444
141 3.198611e-03 0.000 0 0.491769547 0.13043478 0.20000000
142 1.823449e-02 0.000 0 0.491769547 0.13043478 0.20888889
143 3.725677e-02 0.000 1 1.000000000 0.17391304 0.18666667
144 4.598275e-02 0.000 0 1.000000000 0.17391304 0.23555556
145 3.117257e-02 0.000 0 1.000000000 0.17391304 0.15111111
146 2.667217e-02 0.000 0 1.000000000 0.17391304 0.19555556
147 2.415121e-02 0.000 0 1.000000000 0.17391304 0.23555556
148 2.655168e-02 0.000 0 1.000000000 0.17391304 0.21333333
149 2.612873e-02 0.000 0 1.000000000 0.17391304 0.28444444
150 3.065813e-02 0.000 0 1.000000000 0.17391304 0.23111111
151 1.854875e-02 0.000 0 1.000000000 0.17391304 0.36666667
152 1.674724e-02 0.000 0 1.000000000 0.17391304 0.32444444
153 1.259145e-02 0.000 1 1.000000000 0.17391304 0.22888889
154 2.408523e-02 0.000 0 1.000000000 0.17391304 0.32000000
155 1.582030e-02 0.000 1 1.000000000 0.17391304 0.26666667
156 3.966162e-02 0.000 1 1.000000000 0.17391304 0.23555556
157 2.742906e-02 0.000 0 1.000000000 0.17391304 0.18000000
158 1.368171e-02 0.000 0 0.452674897 0.17391304 0.80666667
159 1.502216e-02 0.000 0 0.452674897 0.17391304 0.42888889
160 1.594585e-02 0.000 0 1.000000000 0.17391304 0.40666667
161 1.424235e-02 0.000 1 0.452674897 0.17391304 0.48888889
162 1.637678e-02 0.000 0 0.452674897 0.17391304 1.00000000
163 2.054010e-02 0.000 1 0.452674897 0.17391304 1.00000000
164 1.700238e-02 0.000 1 0.452674897 0.17391304 1.00000000
165 2.513255e-02 0.000 0 0.452674897 0.17391304 0.39333333
166 3.279402e-02 0.000 0 0.452674897 0.17391304 0.44444444
[ reached getOption("max.print") -- omitted 340 rows ]
attr(,"scaled:center")
crim zn chas nox rad medv
0.00632 0.00000 0.00000 0.38500 1.00000 5.00000
attr(,"scaled:scale")
crim zn chas nox rad medv
88.96988 100.00000 1.00000 0.48600 23.00000 45.00000
> {
+ # 학습용과 테스트용 데이터 분리
+ n <- nrow(data.scaled)
+ n
+
+ set.seed(1234)
+ index <- sample(1:n, round(0.8*n))
+ index
+
+ train <- as.data.frame(data.scaled[index,])
+ head(train)
+
+ test <- as.data.frame(data.scaled[-index,])
+ head(test)
+ }
crim zn chas nox rad medv
1 0.0000000000 0.180 0 0.3148148 0.00000000 0.4222222
5 0.0007050701 0.000 0 0.1502058 0.08695652 0.6933333
7 0.0009213230 0.125 0 0.2860082 0.17391304 0.3977778
8 0.0015536719 0.125 0 0.2860082 0.17391304 0.4911111
9 0.0023032514 0.125 0 0.2860082 0.17391304 0.2555556
11 0.0024566741 0.125 0 0.2860082 0.17391304 0.2222222
> {
+ # 학습용 입출력 변수 할당
+ names.col <- colnames(train)
+ names.col
+
+ var.dependent <- names.col[6]
+ var.dependent
+
+ var.independent <- names.col[-6]
+ var.independent
+
+ f.var.independent <- paste(var.independent, collapse = " + ")
+ f.var.independent
+
+ f <- paste(var.dependent, "~", f.var.independent)
+ f
+ }
[1] "medv ~ crim + zn + chas + nox + rad"
> {
+ # 학습 및 모형 출력
+ model <- neuralnet(f,
+ data = train,
+ hidden = c(5, 3, 3, 2),
+ linear.output = T)
+ plot(model)
+ }
> {
+ # 테스트
+ predicted <- compute(model, test)
+ predicted$net.result
+
+ predicted.real <- predicted$net.result * (maxs[6] - mins[6]) + mins[6]
+ predicted.real
+
+ test.real <- test$medv * (maxs[6] - mins[6]) + mins[6]
+ test.real
+ }
[1] 24.0 36.2 22.9 27.1 16.5 15.0 15.2 15.6 16.6 14.5 20.0 24.7
[13] 21.2 19.3 16.6 19.7 18.9 35.4 17.4 28.0 23.9 19.8 19.4 22.8
[25] 18.8 20.4 22.0 16.2 18.0 15.6 13.3 15.6 19.4 13.1 50.0 50.0
[37] 23.1 29.6 36.4 33.3 22.6 23.3 21.5 21.7 50.0 46.7 31.7 23.7
[49] 23.7 18.5 29.6 21.9 36.0 43.5 24.4 33.1 35.1 20.1 24.8 22.0
[61] 28.4 22.8 16.1 22.1 17.8 21.0 18.5 19.3 19.8 19.4 19.0 18.7
[73] 22.9 18.6 22.7 23.1 50.0 50.0 13.3 23.2 11.9 15.0 9.5 16.1
[85] 14.3 9.6 17.1 18.4 15.4 14.1 21.4 19.0 13.3 14.6 23.0 23.7
[97] 21.8 21.2 19.1 24.5 22.4
> {
+ # 실제값 대비 예측값 비교(분포)
+ plot(test.real, predicted.real,
+ xlim=c(0, 50), ylim=c(0, 50),
+ main="실제값 대비 예측값 분포",
+ xlab="실제값", ylab="예측값",
+ col="red",
+ pch=18, cex=0.7)
+ abline(0, 1, col="blue", lty=2)
+
+ MAPE.model <- sum(abs(test.real - predicted.real)/test.real * 100) / length(test.real)
+ MAPE.model
+ }
[1] 20.09686
2.
######################################################
# 패키지 설치 및 로딩
# xlsx 패키지를 위해 먼저 JAVA 구축여부 학인(부록 참조)
install.packages("xlsx")
install.packages("nnet")
library(xlsx)
library(nnet)
# 시계열 데이터 읽기
data <- read.xlsx2(file.choose(), 1)
data
data$종가 <- gsub(",", "", data$종가)
data$종가 <- as.numeric(data$종가)
df <- data.frame(일자=data$년.월.일, 종가=data$종가)
df <- df[order(df$일자), ]
df
n <- nrow(df)
n
rownames(df) <- 1:n
# 데이터 정규화
norm <- (df$종가 - min(df$종가)) / (max(df$종가) - min(df$종가)) * 0.9 + 0.05
norm
df <- cbind(df, 종가norm=norm)
df
# 학습 데이터와 테스트 데이터 분리
n80 <- round(n * 0.8, 0)
n80
df.learning <- df[1:n80, ]
df.learning
df.test <- df[(n80+1):n, ]
df.test
# 인공신경망 구조
INPUT_NODES <- 10
HIDDEN_NODES <- 10
OUTPUT_NODES <- 5
ITERATION <- 100
# 입출력 데이터 파일 생성 함수
getDataSet <- function(item, from, to, size) {
dataframe <- NULL
to <- to - size + 1 # 마지막 행의 시작날짜 번호
for(i in from:to) { # 각 행의 날짜 시작번호에 대한 반복
start <- i # 각 행의 시작날짜 번호
end <- start + size - 1 # 각 행의 끝날짜 번호
temp <- item[c(start:end)] # item에서 start~end 구간의 데이터 추출
dataframe <- rbind(dataframe, t(temp)) # t() 함수를 사용하여 열 단위의 데이터를 행으로 전환 후 데이터 프레임에 추가
}
return(dataframe) # 데이터 프레임 반환
}
# 입력 데이터 만들기
in_learning <- getDataSet(df.learning$종가norm, 1, 92, INPUT_NODES)
in_learning
# 목표치 데이터 만들기
out_learning <- getDataSet(df.learning$종가norm, 11, 97, OUTPUT_NODES)
out_learning
# 학습
model <- nnet(in_learning, out_learning, size=HIDDEN_NODES, maxit=ITERATION)
# 입력 데이터 만들기
in_test <- getDataSet(df.test$종가norm, 1, 19, INPUT_NODES)
in_test
# 예측
predicted_values <- predict(model, in_test, type="raw")
Vpredicted <- (predicted_values - 0.05) / 0.9 * (max(df$종가) - min(df$종가)) + min(df$종가)
Vpredicted
# 예측 기간의 실제 값 데이터 만들기
Vreal <- getDataSet(df.test$종가, 11, 24, OUTPUT_NODES)
Vreal
# 오차 계산 및 출력
ERROR <- abs(Vreal - Vpredicted)
MAPE <- rowMeans(ERROR / Vreal) * 100
MAPE
mean(MAPE)
# 예측 입력 데이터 만들기
in_forecasting <- df$종가norm[112:121]
in_forecasting
# 예측
predicted_values <- predict(model, in_forecasting, type="raw")
Vpredicted <- (predicted_values - 0.05) / 0.9 * (max(df$종가) - min(df$종가)) + min(df$종가)
Vpredicted
# 그래프 출력
plot(82:121, df$종가[82:121], xlab="기간", ylab="종가", xlim=c(82,126), ylim=c(1800, 2100), type="o")
lines(122:126, Vpredicted, type="o", col="red")
abline(v=121, col="blue", lty=2)
#####################################################################
## 13.4 붓꽃 종의 분류(분류문제)
# 데이터 확인
install.packages("neuralnet")
library(neuralnet)
iris
data <- iris
# 결측치 확인
na <- apply(data, 2, is.na)
na
apply(na, 2, sum)
# 데이터 정규화
maxs <- apply(data[,1:4], 2, max)
maxs
mins <- apply(data[,1:4], 2, min)
mins
data[,1:4] <- scale(data[,1:4], center = mins, scale = maxs - mins)
data[,1:4]
summary(data)
# 출력 데이터 생성
data$setosa <- ifelse(data$Species == "setosa", 1, 0)
data$virginica <- ifelse(data$Species == "virginica", 1, 0)
data$versicolor <- ifelse(data$Species == "versicolor", 1, 0)
head(data)
# 학습용과 테스트용 데이터 분리
n <- nrow(data)
n
set.seed(2000)
index <- sample(1:n, round(0.8*n))
index
train <- as.data.frame(data[index,])
head(train)
test <- as.data.frame(data[-index,])
head(test)
# 학습용 입출력 데이터
f.var.independent <- "Sepal.Length + Sepal.Width + Petal.Length + Petal.Width"
f.var.dependent <- "setosa + versicolor + virginica"
f <- paste(f.var.dependent, "~", f.var.independent)
f
# 학습 및 모형 출력
nn <- neuralnet(f, data=train, hidden=c(6, 8, 14, 8, 6), linear.output=F)
plot(nn)
# 테스트
predicted <- compute(nn, test[,-5:-8])
predicted$net.result
idx <- max.col(predicted$net.result)
idx
species <- c('setosa', 'versicolor', 'virginica')
predicted <- species[idx]
predicted
# 실제 값 대비 예측 값 비교(분포)
table(test$Species, predicted)
t <- table(test$Species, predicted)
tot <- sum(t)
tot
same <- sum(diag(t))
same
same / tot
######################################################
install.packages("neuralnet")
install.packages("Metrics")
install.packages("MASS")
library(neuralnet)
library(Metrics)
library(MASS)
data("Boston")
data<-Boston
keeps<-c("crim","indus","nox","rm","age","dis","tax","ptratio","lstat","medv")
data<-data[keeps]
data<-scale(data)
apply(data,2,function(x) sum(is.na(x)))
set.seed(2016)
n = nrow(data)
train <- sample(1:n, 400,FALSE)
f<-medv~ crim+indus + nox + rm + age + dis + tax + ptratio + lstat
fit<-neuralnet(f,
data = data[train,],
hidden = c(10,12,20),
algorithm = "rprop+",
err.fct ="sse",
act.fct = "logistic",
threshold = 0.1,
linear.output = TRUE)
pred<-compute(fit, data[-train, 1:9])
plot(data[-train,10]~pred$net.result)
round(cor(pred$net.result, data[-train,10]),2)
mse(pred$net.result, data[-train,10])
rmse(pred$net.result, data[-train,10])
##############################################
require(deepnet)
train
X = data[train,1:9]
Y = data[train, 10]
fitB <- nn.train(x=X,y=Y,
initW = NULL,
initB = NULL,
hidden = c(10,12,20),
learningrate = 0.58,
momentum = 0.74,
learningrate_scale = 1,
activationfun = "sigm",
output = "linear",
numepochs = 970,
batchsize = 60,
hidden_dropout = 0,
visible_dropout = 0)
predB<-nn.predict(fitB, data[-train,1:9])
round(cor(predB, data[-train,10]),2)
mse(predB, data[-train,10])
rmse(predB, data[-train,10])
#############################
#stepmax=1e6,
#print(
install.packages("neuralnet")
library(neuralnet)
data <- iris
# 결측치 확인
na <- apply(data, 2, is.na)
apply(na, 2, sum)
# 데이터 정규화
maxs <- apply(data[,1:4], 2, max)
mins <- apply(data[,1:4], 2, min)
data[,1:4] <- scale(data[,1:4], center = mins, scale = maxs - mins)
summary(data)
# 출력 데이터 생성
data$setosa <- ifelse(data$Species == "setosa", 1, 0)
data$virginica <- ifelse(data$Species == "virginica", 1, 0)
data$versicolor <- ifelse(data$Species == "versicolor", 1, 0)
head(data)
# 학습용과 테스트용 데이터 분리
n <- nrow(data)
set.seed(2000)
index <- sample(1:n, round(0.8*n))
head(index)
species <- c('setosa', 'versicolor', 'virginica')
train <- as.data.frame(data[index,])
test <- as.data.frame(data[-index,])
# 학습용 입출력 데이터
f.var.independent <- "Sepal.Length + Sepal.Width + Petal.Length + Petal.Width"
f.var.dependent <- "setosa + versicolor + virginica"
f <- paste(f.var.dependent, "~", f.var.independent)
accuracy_k <- NULL
maxCount <- 5
# kk가 1부터 train 행 수까지 증가할 때 (반복문)
for(kk in 1:maxCount){
getIndex1 <- sample(1:maxCount, 1, replace=TRUE)
getIndex1
getIndex2 <- sample(1:maxCount, 1, replace=TRUE)
getIndex2
getIndex3 <- sample(1:maxCount, 1, replace=TRUE)
getIndex3
n <- neuralnet(f, data=train, hidden=c(getIndex1, getIndex2, getIndex3), stepmax=1e6,linear.output=F)
#plot(n)
predicted <- compute(nn, test[,-5:-8])
print(head(predicted$net.result))
idx <- max.col(predicted$net.result)
print(idx)
predicted <- species[idx]
print(predicted)
idx <- max.col(predicted$net.result)
predicted <- species[idx]
print(head(predicted))
# 실제 값 대비 예측 값 비교(분포)
table(test$Species, predicted)
t <- table(test$Species, predicted)
tot <- sum(t)
same <- sum(diag(t))
same / tot
accuracy_k <- c(accuracy_k, (same / tot) * 100)
}
valid_k <- data.frame(k = c(1:maxCount), accuracy = accuracy)
valid_k
# 정확도 그래프 그리기
plot(formula = accuracy ~ k, data = valid_k, type = "o", pch = 20, main = "ANN")예시 일부
library(neuralnet)
> library(Metrics)
Error in library(Metrics) : ‘Metrics’이라고 불리는 패키지가 없습니다
> install.packages("Metrics")
WARNING: Rtools is required to build R packages but is not currently installed. Please download and install the appropriate version of Rtools before proceeding:
https://cran.rstudio.com/bin/windows/Rtools/
‘C:/Users/PC44069/Documents/R/win-library/4.1’의 위치에 패키지(들)을 설치합니다.
(왜냐하면 ‘lib’가 지정되지 않았기 때문입니다)
trying URL 'https://cran.rstudio.com/bin/windows/contrib/4.1/Metrics_0.1.4.zip'
Content type 'application/zip' length 83274 bytes (81 KB)
downloaded 81 KB
package ‘Metrics’ successfully unpacked and MD5 sums checked
The downloaded binary packages are in
C:\Users\Public\Documents\ESTsoft\CreatorTemp\Rtmp6rxTuC\downloaded_packages
> library(MASS)
> library(Metrics)
> data("Boston")
> data<-Boston
> keeps<-c("crim","indus","nox","rm","age","dis","tax","ptratio","lstat","medv")
> data<-data[keeps]
> data<-scale(data)
> apply(data,2,function(x) sum(is.na(x)))
crim indus nox rm age dis tax ptratio
0 0 0 0 0 0 0 0
lstat medv
0 0
> set.seed(2016)
> n = nrow(data)
> train <- sample(1:n, 400,FALSE)
> f<-medv~ crim+indus + nox + rm + age + dis + tax + ptratio + lstat
> fit<-neuralnet(f,
+ data = data[train,],
+ hidden = c(10,12,20),
+ algorithm = "rprop+",
+ err.fct ="sse",
+ act.fct = "logistic",
+ threshold = 0.1,
+ linear.output = TRUE)
> pred<-compute(fit, data[-train, 1:9])
> plot(data[-train,10]~pred$net.result)
> round(cor(pred$net.result, data[-train,10]),2)
[,1]
[1,] 0.91
> mse(pred$net.result, data[-train,10])
[1] 0.1886343
> rmse(pred$net.result, data[-train,10])
[1] 0.4343205
> ##############################################
'R언어' 카테고리의 다른 글
| R언어로 네트워크를 구축해보자 (0) | 2022.05.24 |
|---|---|
| R 스튜디오 테마 바꾸기 (0) | 2022.04.05 |
| R언어 3강 (0) | 2022.03.29 |
| R언어 연산 (0) | 2022.03.22 |
