if(par3!='NA') par3 <- as.numeric(par3) else par3 <- NA
if(par4!='NA') par4 <- as.numeric(par4) else par4 <- NA
par6 <- as.numeric(par6) #Seasonal Period
par9 <- as.numeric(par9) #Forecast Horizon
par10 <- as.numeric(par10) #Alpha
library(forecast)
if (par1 == 'CSV') {
xarr <- read.csv(file=paste('tmp/',par7,'.csv',sep=''),header=T)
numseries <- length(xarr[1,])-1
n <- length(xarr[,1])
nmh <- n - par9
nmhp1 <- nmh + 1
rarr <- array(NA,dim=c(n,numseries))
farr <- array(NA,dim=c(n,numseries))
parr <- array(NA,dim=c(numseries,8))
colnames(parr) = list('ME','RMSE','MAE','MPE','MAPE','MASE','ACF1','TheilU')
for(i in 1:numseries) {
sindex <- i+1
x <- xarr[,sindex]
if(par2=='Croston') {
if (i==1) m <- croston(x,alpha=par10)
if (i==1) mydemand <- m$model$demand[]
fit <- croston(x[1:nmh],h=par9,alpha=par10)
}
if(par2=='ARIMA') {
m <- auto.arima(ts(x,freq=par6),d=par3,D=par4)
mydemand <- forecast(m)
fit <- auto.arima(ts(x[1:nmh],freq=par6),d=par3,D=par4)
}
if(par2=='ETS') {
m <- ets(ts(x,freq=par6),model=par5)
mydemand <- forecast(m)
fit <- ets(ts(x[1:nmh],freq=par6),model=par5)
}
try(rarr[,i] <- mydemand$resid,silent=T)
try(farr[,i] <- mydemand$mean,silent=T)
if (par2!='Croston') parr[i,] <- accuracy(forecast(fit,par9),x[nmhp1:n])
if (par2=='Croston') parr[i,] <- accuracy(fit,x[nmhp1:n])
}
write.csv(farr,file=paste('tmp/',par8,'_f.csv',sep=''))
write.csv(rarr,file=paste('tmp/',par8,'_r.csv',sep=''))
write.csv(parr,file=paste('tmp/',par8,'_p.csv',sep=''))
}
if (par1 == 'Input box') {
numseries <- 1
n <- length(x)
if(par2=='Croston') {
m <- croston(x)
mydemand <- m$model$demand[]
}
if(par2=='ARIMA') {
m <- auto.arima(ts(x,freq=par6),d=par3,D=par4)
mydemand <- forecast(m)
}
if(par2=='ETS') {
m <- ets(ts(x,freq=par6),model=par5)
mydemand <- forecast(m)
}
summary(m)
}
bitmap(file='test1.png')
op <- par(mfrow=c(2,1))
if (par2=='Croston') plot(m)
if ((par2=='ARIMA') | par2=='ETS') plot(forecast(m))
plot(mydemand$resid,type='l',main='Residuals', ylab='residual value', xlab='time')
par(op)
dev.off()
bitmap(file='pic2.png')
op <- par(mfrow=c(2,2))
acf(mydemand$resid, lag.max=n/3, main='Residual ACF', ylab='autocorrelation', xlab='time lag')
pacf(mydemand$resid,lag.max=n/3, main='Residual PACF', ylab='partial autocorrelation', xlab='time lag')
cpgram(mydemand$resid, main='Cumulative Periodogram of Residuals')
qqnorm(mydemand$resid); qqline(mydemand$resid, col=2)
par(op)
dev.off()
load(file='createtable')
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'Demand Forecast',6,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,'Point',header=TRUE)
a<-table.element(a,'Forecast',header=TRUE)
a<-table.element(a,'95% LB',header=TRUE)
a<-table.element(a,'80% LB',header=TRUE)
a<-table.element(a,'80% UB',header=TRUE)
a<-table.element(a,'95% UB',header=TRUE)
a<-table.row.end(a)
for (i in 1:length(mydemand$mean)) {
a<-table.row.start(a)
a<-table.element(a,i+n,header=TRUE)
a<-table.element(a,as.numeric(mydemand$mean[i]))
a<-table.element(a,as.numeric(mydemand$lower[i,2]))
a<-table.element(a,as.numeric(mydemand$lower[i,1]))
a<-table.element(a,as.numeric(mydemand$upper[i,1]))
a<-table.element(a,as.numeric(mydemand$upper[i,2]))
a<-table.row.end(a)
}
a<-table.end(a)
table.save(a,file='mytable.tab')
a<-table.start()
a<-table.row.start(a)
a<-table.element(a,'What is next?',1,TRUE)
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,hyperlink(paste('https://automated.biganalytics.eu/Patrick.Wessa/rwasp_demand_forecasting_simulate.wasp',sep=''),'Simulate Time Series','',target=''))
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,hyperlink(paste('https://automated.biganalytics.eu/Patrick.Wessa/rwasp_demand_forecasting_croston.wasp#output',sep=''),'Generate Forecasts','',target=''))
a<-table.row.end(a)
a<-table.row.start(a)
a<-table.element(a,hyperlink(paste('https://automated.biganalytics.eu/Patrick.Wessa/rwasp_demand_forecasting_analysis.wasp',sep=''),'Forecast Analysis','',target=''))
a<-table.row.end(a)
a<-table.end(a)
table.save(a,file='mytable0.tab')
-SERVER-wessa.org
|
