R language book learning 03 "in simple terms R language data analysis" - Chapter 12 support vector machine Chapter 13 neural network

1 Support vector machine

1.1 summary

Support vector machine （Surpport Vector Machine, SVM) It is a supervised algorithm in machine learning , It can be used to solve classification and regression problems . Support vector machine maps data into points in space , This mapping should form a wide enough interval , Separate the data .

About support vector machines , Be sure to understand the concept of hyperplane , The other is nuclear technology .

1.2 Code implementation

The data in the book is very good , Here we quote （melbench In bag Glass Data sets , These include 6 The chemical element content of three kinds of glass materials ）

library(mlbench)
data(Glass)

smp.size = floor(0.8*nrow(Glass)) 
set.seed(1)                     
train.ind = sample(seq_len(nrow(Glass)), smp.size)
train = Glass[train.ind, ] # 80% Training set of 
test = Glass[-train.ind, ] # 20% Test set of 



library(e1071)
model = svm(formula = Type ~ .,data = train)
summary(model)

Use caret In bag confusionMatrix() Function calculation index . about e1071 Parameter ,cost Parameters determine the soft boundary of support vector machine , Used to punish the wrong samples ,epsilon Parameters mainly affect SVR, Create a tolerance zone , The larger the tolerance range , More data will be ignored .gamma Is a parameter related to the kernel function , Mainly polynomial, radial asis as well as sigmoid.

2 neural network

2.1 summary

In machine learning and cognitive science , Artificial neural network is a statistical model inspired by biology , Used to estimate or approximate a function .

Including the input 、 Hide and output three levels .

Neuron is the basic unit of information processing , Will receive input , To produce output . The output of the neural network generates the output through the weighted sum of the neuron input by the activation function , Different activation functions have different functions . stay R In language nnet\RSNNS\neuralnet Isochronal neural network packet .

2.2 R Language implementation

Use neuralnet Build a neural network model .

n <- names(shuttleTrain)
form <- as.formula(paste("use~", paste(n[!n %in% "use"], collapse = "+")))

fit <- neuralnet(form, 
                 data = shuttleTrain,
                 err.fct = "ce",
                 linear.output = FALSE
)  

Use scorecard In bag perf_eva() Function to draw ROC and KS curve .

3 Other chapters

Chapter 14 is text mining , Involving participle 、 Emotional analysis and LDA The content such as .

Chapter 15 is social network analysis , Including community discovery methods .

Chapter 16 introduces H20 And its learning platform .

Chapter 17 R Language crawler .

Not as the main content of this study .

The end of the book