Chapter 13 Neural Network2 (W12)
뉴런: 신경세포
Figure 13.1: 뉴런
인공신경망
Figure 13.2: 인공신경망
13.1 인공신경망이란? (Artificial Neural Network?)
- 인간의 뇌 기능을 모방한 네트워크
- 인간의 사고방식을 컴퓨터에 학습시키는 것
13.1.1 Input Layer: 입력계층
- 자극 부분
- x1, x2, x3…
13.1.3 출력계층: Output Layer
- 반응 부분
- y1, y2, y3…
- 은닉층의 모든 단위들과 연결
- 출력은 어떻게 결정하는가?
- 활성함수 (activation function)
- 계단 함수, 부호함수,
- 시그모이드 함수
- 선형함수
- 활성함수 (activation function)
13.2 경영학과 인공신경망
- 재무, 회계, 마케팅 등과 같은 다양한 부문에 활용가능
- 인공신경망 모델에서 결정된 가중치를 기반으로 의사결정에 도움을 줌
13.3 예제
- neural network package
- nnet
- neuralnet
- nnet: neural network package
- caret
- 지난번에 배운 의사결정나무 (Decision Tree)에서도 사용했음
- 훈련데이터, 테스트 데이터를 나누기 위해 –> seperate data (training: test)
- sys timezone 에러가 날 경우
- Sys.setenv(TZ=‘GMT’)
13.3.1 nnet 패키지
## ================================================================
## 01. 환경설정 및 데이터 불러오기
## ================================================================
library(nnet)
data(iris)
# nnet 종속변수 입력을 위한 종속변수 변환
species.ind <- class.ind(iris$Species)
iris <- cbind(iris, species.ind)
## ================================================================
## 02. 훈련 및 검증 데이터 분리
## ================================================================
train.idx <- sample(1:150, 100)
iris.train.df <- iris[train.idx, ]
iris.test.df <- iris[-train.idx, ]
## ================================================================
## 03. 신경망 모형 적합
## ================================================================
iris.nn <- nnet(x = iris.train.df[, c(1:4)], y = iris.train.df[,
c(6:8)], size = 10, softmax = TRUE)## # weights: 83
## initial value 120.965425
## iter 10 value 47.031199
## iter 20 value 11.742304
## iter 30 value 6.657025
## iter 40 value 4.812066
## iter 50 value 4.585421
## iter 60 value 4.570077
## iter 70 value 4.569501
## final value 4.569500
## converged## ================================================================
## 04. 검증 데이터 예측 및 모형 평가
## ================================================================
## 훈련데이터
iris.pred <- predict(iris.nn, iris.train.df[, c(1:4)], type = "class")
table(iris.pred, iris.train.df$Species)##
## iris.pred setosa versicolor virginica
## setosa 32 0 0
## versicolor 0 32 0
## virginica 0 1 35# 검증데이터
iris.pred <- predict(iris.nn, iris.test.df[, c(1:4)], type = "class")
table(iris.pred, iris.test.df$Species)##
## iris.pred setosa versicolor virginica
## setosa 18 1 0
## versicolor 0 16 0
## virginica 0 0 15## ================================================================
## 05. 신경망 시각화
## ================================================================
## 시각화 R 코드 함수 다운로드
library(devtools)
source_url("https://gist.githubusercontent.com/fawda123/7471137/raw/466c1474d0a505ff044412703516c34f1a4684a5/nnet_plot_update.r")## SHA-1 hash of file is 74c80bd5ddbc17ab3ae5ece9c0ed9beb612e87ef# 신경망 모형 시각화
library(reshape2)
plot.nnet(iris.nn)## Loading required package: scales##
## Attaching package: 'scales'## The following object is masked from 'package:plotrix':
##
## rescale## Loading required package: reshape##
## Attaching package: 'reshape'## The following objects are masked from 'package:reshape2':
##
## colsplit, melt, recast## The following objects are masked from 'package:plyr':
##
## rename, round_any
검증
# table(iris.pred, iris.train.df$Species)
table(iris.pred, iris.test.df$Species)##
## iris.pred setosa versicolor virginica
## setosa 18 1 0
## versicolor 0 16 0
## virginica 0 0 1513.3.2 neuralnet 패키지
library(neuralnet)
data(iris)
summary(iris)## Sepal.Length Sepal.Width Petal.Length Petal.Width
## Min. :4.300 Min. :2.000 Min. :1.000 Min. :0.100
## 1st Qu.:5.100 1st Qu.:2.800 1st Qu.:1.600 1st Qu.:0.300
## Median :5.800 Median :3.000 Median :4.350 Median :1.300
## Mean :5.843 Mean :3.057 Mean :3.758 Mean :1.199
## 3rd Qu.:6.400 3rd Qu.:3.300 3rd Qu.:5.100 3rd Qu.:1.800
## Max. :7.900 Max. :4.400 Max. :6.900 Max. :2.500
## Species
## setosa :50
## versicolor:50
## virginica :50
##
##
## # neuralnet은 식에서 '.'을 지원하지 않기 때문에 아래와
# 같이 식을 문자열로 생성.
formula <- as.formula(paste("Species ~ ", paste(names(iris)[-length(iris)],
collapse = "+")))
formula## Species ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Widthformula <- as.formula(paste("setosa + versicolor + virginica ~ ",
paste(names(iris)[-length(iris)], collapse = "+")))
formula## setosa + versicolor + virginica ~ Sepal.Length + Sepal.Width +
## Petal.Length + Petal.WidthtrainData <- cbind(iris[, 1:4], class.ind(iris$Species))
head(trainData)## Sepal.Length Sepal.Width Petal.Length Petal.Width setosa versicolor
## 1 5.1 3.5 1.4 0.2 1 0
## 2 4.9 3.0 1.4 0.2 1 0
## 3 4.7 3.2 1.3 0.2 1 0
## 4 4.6 3.1 1.5 0.2 1 0
## 5 5.0 3.6 1.4 0.2 1 0
## 6 5.4 3.9 1.7 0.4 1 0
## virginica
## 1 0
## 2 0
## 3 0
## 4 0
## 5 0
## 6 0multi_model <- neuralnet(formula, trainData, hidden = 3)
plot(multi_model)
species_prob = compute(multi_model, iris[, 1:4])$net.result
species_prob## [,1] [,2] [,3]
## [1,] 1.000057707030813 -0.000102754014119 0.000002342043184
## [2,] 1.000037412323813 -0.000061633125555 0.000008659995948
## [3,] 1.000050211387514 -0.000090805722475 0.000003675827738
## [4,] 1.000029471517131 -0.000065298309382 0.000005035478588
## [5,] 1.000058304962745 -0.000104400431052 0.000002021686664
## [6,] 1.000050100558360 -0.000092784400015 0.000003030382018
## [7,] 1.000037425790689 -0.000079237385818 0.000003231320727
## [8,] 1.000054338770845 -0.000097637430845 0.000002863461942
## [9,] 0.999997938497114 -0.000022318167105 0.000008398418352
## [10,] 1.000053680035691 -0.000095295296346 0.000003379433893
## [11,] 1.000059218844035 -0.000105445293119 0.000001986185131
## [12,] 1.000049961197867 -0.000094033847198 0.000002601027504
## [13,] 1.000052226790819 -0.000091948592202 0.000003955954993
## [14,] 1.000054316344174 -0.000098092819119 0.000002715881275
## [15,] 1.000060298020486 -0.000107616785790 0.000001654893012
## [16,] 1.000059806415190 -0.000106881336328 0.000001727499421
## [17,] 1.000057378519864 -0.000101797972362 0.000002533938474
## [18,] 1.000051821575150 -0.000092357074662 0.000003702655258
## [19,] 1.000057155351855 -0.000101088837022 0.000002682712656
## [20,] 1.000056678979147 -0.000102440903625 0.000002115867441
## [21,] 1.000053637350369 -0.000094078425362 0.000003741573026
## [22,] 1.000043086412591 -0.000082730072325 0.000003930853700
## [23,] 1.000059380217210 -0.000106256561570 0.000001786487914
## [24,] 0.999592883825390 0.000540501137155 0.000054905599752
## [25,] 1.000034907668944 -0.000080776275748 0.000001965715181
## [26,] 1.000022736141138 -0.000037830137765 0.000011397622130
## [27,] 0.999977065066813 0.000008920808075 0.000011484966355
## [28,] 1.000057172128502 -0.000101688013358 0.000002503066869
## [29,] 1.000056962245505 -0.000100345467885 0.000002851491122
## [30,] 1.000035102280847 -0.000074269764420 0.000004034816905
## [31,] 1.000027438148908 -0.000058366833165 0.000006536149678
## [32,] 1.000023275658471 -0.000028069311139 0.000014579347364
## [33,] 1.000060275416487 -0.000107750805822 0.000001606435085
## [34,] 1.000060265219336 -0.000107685576373 0.000001623363862
## [35,] 1.000038933082996 -0.000069873503273 0.000006594422835
## [36,] 1.000055352279090 -0.000095809610881 0.000003745790085
## [37,] 1.000059267968918 -0.000104423080066 0.000002317078474
## [38,] 1.000059643397744 -0.000106702204045 0.000001731594894
## [39,] 1.000028923570512 -0.000063188799494 0.000005514446887
## [40,] 1.000055053250196 -0.000098028615726 0.000002967081972
## [41,] 1.000053250320202 -0.000095021081533 0.000003329130536
## [42,] 0.998780454687770 0.002401379945875 0.000374096245729
## [43,] 1.000045791759843 -0.000088771203913 0.000002915954326
## [44,] 0.999487425706844 0.000583682660142 0.000035118387509
## [45,] 1.000021526887765 -0.000067613448858 0.000001826402168
## [46,] 0.999976785329683 0.000037179348204 0.000020118082768
## [47,] 1.000058806136191 -0.000105637065185 0.000001797412520
## [48,] 1.000044809836466 -0.000085647489537 0.000003571651682
## [49,] 1.000059056791782 -0.000105345409407 0.000001966126532
## [50,] 1.000053896431935 -0.000095406339046 0.000003413112564
## [51,] -0.000210150987537 1.004342208882166 -0.003783363316744
## [52,] -0.000245849622204 1.003876575568043 -0.003237441749020
## [53,] -0.000250882124698 0.997967514013065 0.002707147356275
## [54,] -0.000250909942631 1.003661303450536 -0.002988756465711
## [55,] -0.000252602027148 0.995991187995093 0.004691757319344
## [56,] -0.000248964016898 1.002923042899643 -0.002301751897053
## [57,] -0.000253201863899 0.996481590975635 0.004192134497870
## [58,] 0.000973052133071 1.001214667555427 -0.004687719026622
## [59,] -0.000218162953650 1.004265410165551 -0.003694788476942
## [60,] -0.000251113183953 1.004160301714881 -0.003514634413817
## [61,] -0.000089535983083 1.004264770146667 -0.004069444061973
## [62,] -0.000250421407962 1.003908828739348 -0.003251723901347
## [63,] 0.000019741427792 1.004213915450077 -0.004046200051713
## [64,] -0.000251905424107 0.998152941679787 0.002512097788502
## [65,] -0.000018355339316 1.003992665140875 -0.004138344469015
## [66,] -0.000201370287834 1.004467668236280 -0.003929411806198
## [67,] -0.000253577056647 0.994130414688387 0.006540963346597
## [68,] 0.000754803291637 1.001824689022496 -0.004564439882854
## [69,] -0.000220717585274 0.869791335308692 0.130788723622740
## [70,] -0.000043279827724 1.004090815927324 -0.004104328920576
## [71,] -0.000136246102342 0.554071107742939 0.446230825074212
## [72,] -0.000165209701094 1.004446339700435 -0.004014713920653
## [73,] -0.000175059904467 0.699472705277849 0.300956968145960
## [74,] -0.000228987787472 1.003765695310316 -0.003237703115567
## [75,] -0.000179752003878 1.004437184524178 -0.003971172492017
## [76,] -0.000223721662793 1.004423473536461 -0.003826053384413
## [77,] -0.000250032023449 0.999711231687524 0.000964810404128
## [78,] -0.000174715395534 0.698017993021939 0.302410577334285
## [79,] -0.000253282091284 0.997099281710058 0.003579191775218
## [80,] 0.003796546341829 0.997635621635096 -0.004906047881831
## [81,] -0.000080212730023 1.004215158078186 -0.004077265363160
## [82,] 0.000530042095040 1.002910369138904 -0.004301831387640
## [83,] -0.000072046190828 1.004189185088280 -0.004084330521366
## [84,] -0.000036945486011 0.183960291607871 0.816016130994739
## [85,] -0.000252773549113 0.988850354376578 0.011817841421489
## [86,] -0.000254028643886 1.002244663476543 -0.001587150522701
## [87,] -0.000250046353339 1.002045998693082 -0.001376450145517
## [88,] -0.000251154847729 1.001756949094558 -0.001073890448362
## [89,] -0.000197982075018 1.004306572627713 -0.004032357141978
## [90,] -0.000244802805880 1.004266822035936 -0.003628883698454
## [91,] -0.000243839756898 1.003652480853514 -0.003068886117610
## [92,] -0.000249392760835 1.002612099299157 -0.001969487141768
## [93,] -0.000164972857789 1.004404107809636 -0.004016122496769
## [94,] 0.000720295515931 1.002316445221534 -0.004426897321123
## [95,] -0.000242578865183 1.004170323615589 -0.003566217256682
## [96,] -0.000091021068294 1.003825734765755 -0.004186235645601
## [97,] -0.000222056505224 1.004375744583858 -0.003906808830362
## [98,] -0.000199355662396 1.004419957242795 -0.003939685601641
## [99,] 0.001646081125560 1.000912667826840 -0.004569991975673
## [100,] -0.000221674548492 1.004431401459793 -0.003920409049274
## [101,] 0.000014792422715 -0.008958797909592 1.008765877439017
## [102,] 0.000013934992122 -0.005764685981363 1.005574560381948
## [103,] 0.000014774533527 -0.008892078266609 1.008699214702552
## [104,] 0.000013996447124 -0.005997856791354 1.005807494941253
## [105,] 0.000014789623473 -0.008948503657473 1.008755591685660
## [106,] 0.000014792020488 -0.008957312554778 1.008764392969844
## [107,] -0.000019719602928 0.119664745412402 0.880255402830331
## [108,] 0.000014737980409 -0.008755107524581 1.008562351657879
## [109,] 0.000014726413207 -0.008712106700164 1.008519396790981
## [110,] 0.000014791862713 -0.008956743560687 1.008763824756682
## [111,] 0.000003849123831 0.031841431401837 0.968001509629419
## [112,] 0.000014032147449 -0.006123256568666 1.005932818599822
## [113,] 0.000014647771279 -0.008419333555614 1.008226883981801
## [114,] 0.000014651342792 -0.008433206627444 1.008240745962066
## [115,] 0.000014790347368 -0.008951070017416 1.008758156289380
## [116,] 0.000014743703493 -0.008777413281844 1.008584651149024
## [117,] 0.000010992159505 0.005207584616165 0.994611867596643
## [118,] 0.000014788425351 -0.008944941553872 1.008752029597009
## [119,] 0.000014792470362 -0.008958952807339 1.008766032240195
## [120,] -0.000020001002842 0.120835012637147 0.879086113755910
## [121,] 0.000014783347316 -0.008925029337794 1.008732137945863
## [122,] 0.000013815072666 -0.005317804384438 1.005128076983902
## [123,] 0.000014792093475 -0.008957541264220 1.008764621442475
## [124,] -0.000015464342977 0.103868112134571 0.896038212397043
## [125,] 0.000014659805210 -0.008465975227120 1.008273479789199
## [126,] 0.000013332838342 -0.003510780719116 1.003322557368002
## [127,] -0.000052211150519 0.240876380455642 0.759150497745639
## [128,] -0.000057215217866 0.259507807304846 0.740535327390414
## [129,] 0.000014780252029 -0.008913594123537 1.008720712294633
## [130,] -0.000011124131546 0.087871221917161 0.912020505094601
## [131,] 0.000014749347400 -0.008797054776500 1.008604271497993
## [132,] 0.000014213981154 -0.006802001910561 1.006610906748458
## [133,] 0.000014789132341 -0.008946583925286 1.008753673855029
## [134,] -0.000181455175655 0.723443847524644 0.277005398068689
## [135,] -0.000005568463820 0.066925170252298 0.932947677857599
## [136,] 0.000014791294621 -0.008954540963388 1.008761624118726
## [137,] 0.000014788631911 -0.008944851398255 1.008751942730069
## [138,] 0.000009765519975 0.009768437211712 0.990054984025546
## [139,] -0.000084669959829 0.361860826025513 0.638272364035027
## [140,] 0.000014223464374 -0.006836645259283 1.006645585152378
## [141,] 0.000014790478666 -0.008951559406089 1.008758645178157
## [142,] 0.000014470272829 -0.007757421020486 1.007565556642096
## [143,] 0.000013934992122 -0.005764685981363 1.005574560381948
## [144,] 0.000014790526311 -0.008951812281595 1.008758897583230
## [145,] 0.000014791730898 -0.008956226940531 1.008763308671269
## [146,] 0.000014725803014 -0.008710359421544 1.008517656883509
## [147,] 0.000013029311348 -0.002382886055027 1.002195744255834
## [148,] 0.000012630317513 -0.000896985652979 1.000711141857962
## [149,] 0.000014748575329 -0.008796141283085 1.008603361677057
## [150,] -0.000001031329649 0.050006496193439 0.949852306748718