CODEWITHSUNDEEP
rforecasting
Atirag
Atirag

Reputation: 1740

ARIMA time series forecasting not working in R

I'm trying to forecasting on a time series that looks like this (this is testing data)

[1] 1 1 1 1 4 1 3 4 5 6 5 1 1 1 1 1 5 3 4 7 5 5 6 4 1 1 1 1 1 1 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 8 2

I get the data from a US state. Let's say Alabama like this

res <- subset(d, states == 'Alabama', select = c(levels, weeks))

I then turn the levels data into a time series like this.

tsn = ts(res[[1]])

I then get the best arima model fit like this

aa <- auto.arima(tsn)

For which the result is this

Series: tsn 
ARIMA(1,0,0) with non-zero mean 

Coefficients:
         ar1  intercept
      0.4722     2.2833
s.e.  0.1252     0.4644

sigma^2 estimated as 2.989:  log likelihood=-94.51
AIC=195.03   AICc=195.57   BIC=200.64

I then try to use the forecast function like this

forecast(aa)

And this is when I get this error

Error in forecast(aa) : unused argument (aa)

Any idea how to make the forecast work?

Edit to add code

This is how the data looks like

st      URL                         WEBSITE                 al      aln     wk          WEEKSEASON
Alabama http://adph.org/influenza/  Influenza Surveillance  Level 1 Minimal Oct-04-2008 40  2008-09
Alabama http://adph.org/influenza/  Influenza Surveillance  Level 1 Minimal Oct-11-2008 41  2008-09
Alaska  http://adph.org/influenza/  Influenza Surveillance  Level 1 Minimal Oct-18-2008 42  2008-09
Alaska  http://adph.org/influenza/  Influenza Surveillance  Level 1 Minimal Oct-25-2008 43  2008-09

This is how the code looks like

library(forecast)
library(tseries)

#Extracts relevant data from the csv file
extract_data<-function(){

  #open the file. NAME SHOULD BE CHANGED
  sd <- read.csv(file="sdr.csv",head=TRUE,sep=",")

  #Extracts the data from the ACTIVITY LEVEL column. Notice that the name of the column was changed on the file
  #to 'al' to make the reference easier
  lv_list <- sd$al
  #Gets only the number from each value getting rid of the word "Level"
  lvs <- sapply(strsplit(as.character(lv_list), " "), function(x) x[2])

  #Gets the ACTIVITY LEVEL NAME. Column name was changed to 'aln' on the file
  lvn_list <- sd$aln

  #Gets the state. Column name was changed to 'st' on the file
  st_list <- sd$st

  #Gets the week. Column name was changed to 'wk' on the file
  wlist <- sd$wk
  #Divides the weeks data in month, day, year
  wks <- sapply(strsplit(as.character(wlist), "-"), function(x) c(x[1], x[2], x[3]))

  #Creates a data frame with the selected results. You can choose which data is needed.
  result<-data.frame("states"=st_list,"levels"=lvs,"lvlnames"=lvn_list,"weeks"=wlist)  

  return(result)

}

forecast<-function(){

  d=extract_data()

  #Get data from each state
  res <- subset(d, states == 'Alabama', select = c(levels, weeks))

  #turn data into a time series
  tsn = ts(res[[1]])

  #Plot forecast data with ARIMA models (use differenciated data if needed)
  aa <- auto.arima(tsn)
  forecast(aa)

  return(0) #return results
}

Upvotes: 0

Views: 2243

Answers (1)

cjohnson318
cjohnson318

Reputation: 3243

The variable aa is the model estimate for the data d. Use ARIMA(1,0,0) from aa and plug it into forecast.Arima as follows.

f <- forecast.Arima( d, order=c(1,0,0) )

I did some examples a while back on my blog. Good luck!

Upvotes: 2

Related Questions