Combining Linear Regression and Time Series

4

1

Does anyone know of a predictive model that can combine the linear regression model and time series model?

I have some data about some products. The data has two parts, some attributes about the products themselves, and some time series data of month-by-month sales amounts for each product. We want to predict future month-by-month sales amount for each product. with the products attributes data, we can build a regression model, or an svm model, to predict sales amounts. With the time series sales amount, we can build a time series model to predict sales amount, but is there any predictive model that can combine the regression, svm, and time series models to do the prediction?

zezi

Posted 2015-12-09T22:12:27.573

Reputation: 41

One naive solution is to simply combine the time series data over some fixed window (say the last N months) with the immutable product features and train on that feature set. – jamesmf – 2015-12-09T23:03:47.077

found this, panel data analysis:http://imai.princeton.edu/teaching/files/panel.pdf

– zezi – 2015-12-09T23:36:58.493

Answers

1

You can use time as an additional attribute of the data. If you are having monthly data then your time feature could have 12 different values from 0 to 11. If you are having quarterly data your time attribute could have 4 different value from 0 to 3. You can refer Forecasting: Principles and Practice for more clarification. It's a good book.

Sandeep Kumar

Posted 2015-12-09T22:12:27.573

Reputation: 41