Home » date » 2010 » Dec » 25 »

workshop 10 link 4

*The author of this computation has been verified*

R Software Module: /rwasp_regression_trees1.wasp (opens new window with default values)

Title produced by software: Recursive Partitioning (Regression Trees)

Date of computation: Sat, 25 Dec 2010 10:23:17 +0000

Cite this page as follows:

Statistical Computations at FreeStatistics.org, Office for Research Development and Education, URL http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm.htm/, Retrieved Sun, 26 May 2013 05:24:56 +0000

Original text written by user:

IsPrivate?

No (this computation is public)

User-defined keywords:

System-generated keywords (parent):

t1293195798l1aaimmppqhfdgw (pk = 114890)

Estimated Impact

Dataseries X:

» Textfile « » CSV « » Correlation Matrix « » Notched Boxplots «

-5	-6	33	5	15
-1	-3	24	6	17
-2	-4	24	6	13
-5	-7	31	5	12
-4	-7	25	5	13
-6	-7	28	3	10
-2	-3	24	5	14
-2	0	25	5	13
-2	-5	16	5	10
-2	-3	17	3	11
2	3	11	6	12
1	2	12	6	7
-8	-7	39	4	11
-1	-1	19	6	9
1	0	14	5	13
-1	-3	15	4	12
2	4	7	5	5
2	2	12	5	13
1	3	12	4	11
-1	0	14	3	8
-2	-10	9	2	8
-2	-10	8	3	8
-1	-9	4	2	8
-8	-22	7	-1	0
-4	-16	3	0	3
-6	-18	5	-2	0
-3	-14	0	1	-1
-3	-12	-2	-2	-1
-7	-17	6	-2	-4
-9	-23	11	-2	1
-11	-28	9	-6	-1
-13	-31	17	-4	0
-11	-21	21	-2	-1
-9	-19	21	0	6
-17	-22	41	-5	0
-22	-22	57	-4	-3
-25	-25	65	-5	-3
-20	-16	68	-1	4
-24	-22	73	-2	1
-24	-21	71	-4	0
-22	-10	71	-1	-4
-19	-7	70	1	-2
-18	-5	69	1	3
-17	-4	65	-2	2
-11	7	57	1	5
-11	6	57	1	6
-12	3	57	3	6
-10	10	55	3	3
-15	0	65	1	4
-15	-2	65	1	7
-15	-1	64	0	5
-13	2	60	2	6
-8	8	43	2	1
-13	-6	47	-1	3
-9	-4	40	1	6
-7	4	31	0	0
-4	7	27	1	3
-4	3	24	1	4
-2	3	23	3	7
0	8	17	2	6

Output produced by software:

Enter (or paste) a matrix (table) containing all data (time) series. Every column represents a different variable and must be delimited by a space or Tab. Every row represents a period in time (or category) and must be delimited by hard returns. The easiest way to enter data is to copy and paste a block of spreadsheet cells. Please, do not use commas or spaces to seperate groups of digits!

Summary of computational transaction
Raw Input	view raw input (R code)
Raw Output	view raw output of R engine
Computing time	32 seconds
R Server	'George Udny Yule' @ 72.249.76.132

Confusion Matrix (predicted in columns / actuals in rows)
	C1	C2
C1	30	1
C2	7	22

Charts produced by software:

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/29tik1293272563.png (opens in new window)

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/29tik1293272563.ps (opens in new window)

Click here to open pdf file.

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/39tik1293272563.png (opens in new window)

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/39tik1293272563.ps (opens in new window)

Click here to open pdf file.

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/4j20n1293272563.png (opens in new window)

http://www.freestatistics.org/blog/date/2010/Dec/25/t1293272752z4t0akcefmkpsnm/4j20n1293272563.ps (opens in new window)

Click here to open pdf file.

Parameters (Session):

par1 = 3 ; par2 = quantiles ; par3 = 2 ; par4 = no ;

Parameters (R input):

par1 = 3 ; par2 = quantiles ; par3 = 2 ; par4 = no ;

R code (references can be found in the software module):

library(party)
library(Hmisc)
par1 <- as.numeric(par1)
par3 <- as.numeric(par3)
x <- data.frame(t(y))
is.data.frame(x)
x <- x[!is.na(x[,par1]),]
k <- length(x[1,])
n <- length(x[,1])
colnames(x)[par1]
x[,par1]
if (par2 == 'kmeans') {
cl <- kmeans(x[,par1], par3)
print(cl)
clm <- matrix(cbind(cl$centers,1:par3),ncol=2)
clm <- clm[sort.list(clm[,1]),]
for (i in 1:par3) {
cl$cluster[cl$cluster==clm[i,2]] <- paste('C',i,sep='')
}
cl$cluster <- as.factor(cl$cluster)
print(cl$cluster)
x[,par1] <- cl$cluster
}
if (par2 == 'quantiles') {
x[,par1] <- cut2(x[,par1],g=par3)
}
if (par2 == 'hclust') {
hc <- hclust(dist(x[,par1])^2, 'cen')
print(hc)
memb <- cutree(hc, k = par3)
dum <- c(mean(x[memb==1,par1]))
for (i in 2:par3) {
dum <- c(dum, mean(x[memb==i,par1]))
}
hcm <- matrix(cbind(dum,1:par3),ncol=2)
hcm <- hcm[sort.list(hcm[,1]),]
for (i in 1:par3) {
memb[memb==hcm[i,2]] <- paste('C',i,sep='')
}
memb <- as.factor(memb)
print(memb)
x[,par1] <- memb
}
if (par2=='equal') {
ed <- cut(as.numeric(x[,par1]),par3,labels=paste('C',1:par3,sep=''))
x[,par1] <- as.factor(ed)
}
table(x[,par1])
colnames(x)
colnames(x)[par1]
x[,par1]
if (par2 == 'none') {
m <- ctree(as.formula(paste(colnames(x)[par1],' ~ .',sep='')),data = x)
}
load(file='createtable')
if (par2 != 'none') {
m <- ctree(as.formula(paste('as.factor(',colnames(x)[par1],') ~ .',sep='')),data = x)
if (par4=='yes') {
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'10-Fold Cross Validation',3+2*par3,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'',1,TRUE)
a<-table.element(a,'Prediction (training)',par3+1,TRUE)
a<-table.element(a,'Prediction (testing)',par3+1,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'Actual',1,TRUE)
for (jjj in 1:par3) a<-table.element(a,paste('C',jjj,sep=''),1,TRUE)
a<-table.element(a,'CV',1,TRUE)
for (jjj in 1:par3) a<-table.element(a,paste('C',jjj,sep=''),1,TRUE)
a<-table.element(a,'CV',1,TRUE)
a<-table.row.end(a)
for (i in 1:10) {
ind <- sample(2, nrow(x), replace=T, prob=c(0.9,0.1))
m.ct <- ctree(as.formula(paste('as.factor(',colnames(x)[par1],') ~ .',sep='')),data =x[ind==1,])
if (i==1) {
m.ct.i.pred <- predict(m.ct, newdata=x[ind==1,])
m.ct.i.actu <- x[ind==1,par1]
m.ct.x.pred <- predict(m.ct, newdata=x[ind==2,])
m.ct.x.actu <- x[ind==2,par1]
} else {
m.ct.i.pred <- c(m.ct.i.pred,predict(m.ct, newdata=x[ind==1,]))
m.ct.i.actu <- c(m.ct.i.actu,x[ind==1,par1])
m.ct.x.pred <- c(m.ct.x.pred,predict(m.ct, newdata=x[ind==2,]))
m.ct.x.actu <- c(m.ct.x.actu,x[ind==2,par1])
}
}
print(m.ct.i.tab <- table(m.ct.i.actu,m.ct.i.pred))
numer <- 0
for (i in 1:par3) {
print(m.ct.i.tab[i,i] / sum(m.ct.i.tab[i,]))
numer <- numer + m.ct.i.tab[i,i]
}
print(m.ct.i.cp <- numer / sum(m.ct.i.tab))
print(m.ct.x.tab <- table(m.ct.x.actu,m.ct.x.pred))
numer <- 0
for (i in 1:par3) {
print(m.ct.x.tab[i,i] / sum(m.ct.x.tab[i,]))
numer <- numer + m.ct.x.tab[i,i]
}
print(m.ct.x.cp <- numer / sum(m.ct.x.tab))
for (i in 1:par3) {
a<-table.row.start(a)
a<-table.element(a,paste('C',i,sep=''),1,TRUE)
for (jjj in 1:par3) a<-table.element(a,m.ct.i.tab[i,jjj])
a<-table.element(a,round(m.ct.i.tab[i,i]/sum(m.ct.i.tab[i,]),4))
for (jjj in 1:par3) a<-table.element(a,m.ct.x.tab[i,jjj])
a<-table.element(a,round(m.ct.x.tab[i,i]/sum(m.ct.x.tab[i,]),4))
a<-table.row.end(a)
}
a<-table.row.start(a)
a<-table.element(a,'Overall',1,TRUE)
for (jjj in 1:par3) a<-table.element(a,'-')
a<-table.element(a,round(m.ct.i.cp,4))
for (jjj in 1:par3) a<-table.element(a,'-')
a<-table.element(a,round(m.ct.x.cp,4))
a<-table.row.end(a)
a<-table.end(a)
table.save(a,file='mytable3.tab')
}
}
m
bitmap(file='test1.png')
plot(m)
dev.off()
bitmap(file='test1a.png')
plot(x[,par1] ~ as.factor(where(m)),main='Response by Terminal Node',xlab='Terminal Node',ylab='Response')
dev.off()
if (par2 == 'none') {
forec <- predict(m)
result <- as.data.frame(cbind(x[,par1],forec,x[,par1]-forec))
colnames(result) <- c('Actuals','Forecasts','Residuals')
print(result)
}
if (par2 != 'none') {
print(cbind(as.factor(x[,par1]),predict(m)))
myt <- table(as.factor(x[,par1]),predict(m))
print(myt)
}
bitmap(file='test2.png')
if(par2=='none') {
op <- par(mfrow=c(2,2))
plot(density(result$Actuals),main='Kernel Density Plot of Actuals')
plot(density(result$Residuals),main='Kernel Density Plot of Residuals')
plot(result$Forecasts,result$Actuals,main='Actuals versus Predictions',xlab='Predictions',ylab='Actuals')
plot(density(result$Forecasts),main='Kernel Density Plot of Predictions')
par(op)
}
if(par2!='none') {
plot(myt,main='Confusion Matrix',xlab='Actual',ylab='Predicted')
}
dev.off()
if (par2 == 'none') {
detcoef <- cor(result$Forecasts,result$Actuals)
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'Goodness of Fit',2,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'Correlation',1,TRUE)
a<-table.element(a,round(detcoef,4))
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'R-squared',1,TRUE)
a<-table.element(a,round(detcoef*detcoef,4))
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'RMSE',1,TRUE)
a<-table.element(a,round(sqrt(mean((result$Residuals)^2)),4))
a<-table.row.end(a)
a<-table.end(a)
table.save(a,file='mytable1.tab')
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'Actuals, Predictions, and Residuals',4,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'#',header=TRUE)
a<-table.element(a,'Actuals',header=TRUE)
a<-table.element(a,'Forecasts',header=TRUE)
a<-table.element(a,'Residuals',header=TRUE)
a<-table.row.end(a)
for (i in 1:length(result$Actuals)) {
a<-table.row.start(a)
a<-table.element(a,i,header=TRUE)
a<-table.element(a,result$Actuals[i])
a<-table.element(a,result$Forecasts[i])
a<-table.element(a,result$Residuals[i])
a<-table.row.end(a)
}
a<-table.end(a)
table.save(a,file='mytable.tab')
}
if (par2 != 'none') {
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'Confusion Matrix (predicted in columns / actuals in rows)',par3+1,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'',1,TRUE)
for (i in 1:par3) {
a<-table.element(a,paste('C',i,sep=''),1,TRUE)
}
a<-table.row.end(a)
for (i in 1:par3) {
a<-table.row.start(a)
a<-table.element(a,paste('C',i,sep=''),1,TRUE)
for (j in 1:par3) {
a<-table.element(a,myt[i,j])
}
a<-table.row.end(a)
}
a<-table.end(a)
table.save(a,file='mytable2.tab')
}