R深度学习笔记_r语言pmod和imod区别-CSDN博客

本文链接：https://blog.csdn.net/yunini2/article/details/60574902

本文通过使用mlbench库中的Pima Indians Diabetes数据集，构建了两种不同的神经网络模型来预测糖尿病的发生。利用R语言及其相关包RSNNS和AMORE，文章详细记录了数据预处理、模型训练及评估的过程。最终结果显示，两种模型都达到了较为理想的预测效果。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

library(mlbench)
data("PimaIndiansDiabetes2",package = "mlbench")
ncol(PimaIndiansDiabetes2)
nrow(PimaIndiansDiabetes2)
str(PimaIndiansDiabetes2)
sapply(PimaIndiansDiabetes2,function(x) sum(is.na(x)))
temp<-(PimaIndiansDiabetes2)
temp$insulin<-NULL
temp$triceps<-NULL
temp<-na.omit(temp)
nrow(temp)
ncol(temp)
y<-(temp$diabetes)
temp$diabetes<-NULL
temp<-scale(temp)
temp<-cbind(as.factor(y),temp)
class(temp)
summary(temp)
set.seed(2016)
n=nrow(temp)
n_train<-600
n_test<-n-n_train
train<-sample(1:n,n_train,FALSE)
require(RSNNS)
set.seed(2016)
X<-temp[train,1:6]
Y<-temp[train,7]
fitMLP<-mlp(x=X,y=Y,size = c(12,8),maxit = 1000,
            initFunc = "Random_Weights",
            initFuncParams = c(-0.3,0.3),
            learnFunc = "Std_Backpropagation",
            learnFuncParams = c(0.2,0),
            updateFunc = "Topological_Order",
            updateFuncParams = c(0),
            hiddenActFunc = "Act_Logistic",
            shufflePatterns = TRUE,
            linOut = TRUE)
predMLP<-sign(predict(fitMLP,temp[-train,1:6]))
table(predMLP,sign(temp[-train,7]),dnn=c("Predicted","Observed"))

         Observed
Predicted -1  1
       -1 60 13
       1  14 37

error_rate=(1-sum(predMLP==sign(temp[-train,7]))/124)
round(error_rate,3)

[1] 0.218
========================================================================================================

detach("package:RSNNS",unload = TRUE)
library(AMORE)
net<-newff(n.neurons = c(6,12,8,1),
           learning.rate.global = 0.01,
           momentum.global = 0.5,
           error.criterium = "LMLS",
           Stao=NA,
           hidden.layer = "sigmoid",
           output.layer = "purelin",
           method = "ADAPTgdwm")
X<-temp[train,]
Y<-temp[train,7]
fit<-train(net,P=X,T=Y,error.criterium="LMLS",
           report=T,show.step=100,n.shows=5)

index.show: 1 LMLS 0.185166160641345 
index.show: 2 LMLS 0.176012124255052 
index.show: 3 LMLS 0.169291895233891 
index.show: 4 LMLS 0.162183154823355 
index.show: 5 LMLS 0.156368572375286

pred<-sign(sim(fit$net,temp[-train,]))
table(pred,sign(temp[-train,7]),
           dnn=c("Predicted","Observed"))

         Observed
Predicted -1  1
       -1 66 15
       1   8 35

error_rate=(1-sum(pred==sign(temp[-train,7]))/124)
round(error_rate,3)

[1] 0.185

========================================================================================

先做到这里。。。