ETS/ARIMA Time Series Prediction Boost

Question

Hi everyone!

I'm trying to build a prediction model based in only one variable. Its a daily data, where I want to predict the next 6 days.

I have some series and they dont have correlation, so, the workflow need to predict with one series too well like in other series.

The problem here is it, with some series my workflow predict a growing line trend but in others he predicts the same value for all days, just changing the values in confidence interval.

Which configurations I need to look better to boost a TS model? I'm using the ETS because I got AIC higher than with ARIMA.

Example of series:

2.5100002.6600002.6400002.8200003.0000002.9200002.8200002.6400002.7100002.610000

EricWe · Accepted Answer

Hi @marcusblackhill,

You may receive the same value for all forecasted periods if there are not enough time intervals for the algorithm to make a valid prediction.

These Community articles may help.
How to use the ETS tool 
How to use the ARIMA tool

Also, there is a recorded training session on the Time Series tools.
Time Series Modeling

Additional Time Series tools are available from the Public Gallery. 
Public Gallery Predictive District

Please contact support@alteryx.com for any assistance needed.