deep learning with R

最新推荐文章于 2025-09-13 01:38:56 发布

最新推荐文章于 2025-09-13 01:38:56 发布 · 1.3k 阅读

文章标签：

本文通过具体实例介绍了如何使用R语言中的neuralnet包构建并训练神经网络模型。首先创建了一个简单的数据集并用其训练了一个包含两个隐藏层的神经网络，然后展示了如何利用训练好的模型进行预测并与实际值进行对比。此外还提供了加载并使用Boston数据集进行预处理的方法。

require(RSNNS)
library(neuralnet)
set.seed(2016)
attribute<-as.data.frame(sample(seq(-2,2,length=50),50,replace=F),ncol=1)
response<-attribute^2
data<-cbind(attribute,response)
colnames(data)<-c("attribute","response")
head(data,10)

    attribute   response
1  -1.2653061 1.60099958
2  -1.4285714 2.04081633
3   1.2653061 1.60099958
4  -1.5102041 2.28071637
5  -0.2857143 0.08163265
6  -1.5918367 2.53394419
7   0.2040816 0.04164931
8   1.1020408 1.21449396
9  -2.0000000 4.00000000
10 -1.8367347 3.37359434

fit<-neuralnet(response~attribute,data=data,hidden=c(3,3),threshold = 0.01)
testdata<-as.matrix(sample(seq(-2,2,length=10),10,replace=F),ncol=1)
pred<-compute(fit,testdata)
result<-cbind(testdata,pred$net.result,testdata^2)
colnames(result)<-c("Attribute","Prediction","Actual")
round(result,4)

      Attribute Prediction Actual
 [1,]   -1.5556     2.4213 2.4198
 [2,]   -0.2222     0.0364 0.0494
 [3,]   -1.1111     1.2254 1.2346
 [4,]    1.1111     1.2013 1.2346
 [5,]    0.6667     0.4395 0.4444
 [6,]    1.5556     2.4521 2.4198
 [7,]   -0.6667     0.4554 0.4444
 [8,]    0.2222     0.0785 0.0494
 [9,]    2.0000     3.9317 4.0000
[10,]   -2.0000     3.9675 4.0000

require(Metrics)
data("Boston",package="MASS")
data<-Boston
keeps<-c("crim","indus","nox","rm","age","dis","tax","ptratio","lstat","medv")
data<-data[keeps]
apply(data,2,function(x) sum(is.na(x)))

  crim   indus     nox      rm     age     dis     tax ptratio   lstat 
      0       0       0       0       0       0       0       0       0 
   medv 
      0