Forecasting stationary data that has no trend/seasonality

  • Thread starter Deimantas
  • Start date
  • #1
41
0

Homework Statement



We've got a random variable that appears to have no trend/seasonality, is stationary, and we want to forecast it.
The variable is number of warranty claims received each day, 53 days, so we've got 53 values, and we want to forecast the values of the upcoming 5 days.

2. The attempt at a solution

I'm trying to model the data using ARIMA model.
1.png


Judging from the autocorrelation plots, the data is stationary, so no differencing should be done. Judging from the ACF and PACF plots, our best bet would be AR(1), MA(1) or ARIMA(1,0,1). All yield similar results:



results1.png
results2.png
forecast.png


Is there no better way to forecast this variable? ARIMA does not seem like a good forecasting option in this case. The data has no apparent trends/seasonalities and is stationary. What method would be best to forecast such data?
 

Answers and Replies

  • #2
FactChecker
Science Advisor
Gold Member
6,266
2,434
ARIMA is a very powerful and general method for modeling time series. I don't know what you might try that would be better.
 

Related Threads on Forecasting stationary data that has no trend/seasonality

Replies
2
Views
814
Replies
8
Views
6K
  • Last Post
Replies
5
Views
3K
Replies
6
Views
7K
Replies
11
Views
5K
Replies
3
Views
309
Replies
23
Views
6K
Replies
3
Views
5K
Replies
13
Views
292
Replies
3
Views
794
Top