下面显示了四种预测时间序列的方法。
支持向量机（R package e1071。“Chih-Chung Chang and Chih-Jen Lin，LIBSVM：a library for support vector machines，2005.”的实现）。
递归分区（R package rpart。“Breiman，Friedman，Olshen and Stone。Classification and Regression Trees，1984”的实现）。
将最后两种方法的性能与rle进行比较，得到svm的95％和rpart的94％。
R ：
# Apply rle (forward and backward) and a condition: lenght time for sleep changes of 1h m$rle(Xvar ='sleep',Xlmin =60)m$setZoo()# Show differences between conditional and conditional + rleplot(m$zo[,c(5,7,8)],type ='l')
# Subset a week
# Plot correlation matrix
w$correlation(Xvars =w$nm[c(2:7,9)])
# SVM and Recursive partitioning
plot(tune.gamma.cost)
rpart.p <- predict(rpart.m, data[,-1],type ='class')# Resultsdt$svm = as.integer(svm.p)dt$rpart = as.integer(rpart.p)plot(w$dt2zoo(dt)[,c(5,8,9,10)],type ='l')
▍关注我们