I'm a novice attempting to predict automobile sales using a combination of previous sales (seasonal AR model), macroeconomic indicators such as CPI, consumer sentiment index etc. and more significantly, I'm using weekly data from google search trends for the automobile and other variables from the category (automobile) in google trends. (code developed using Choi and Varian's 2009 paper). The model essentially looks at search trend indices for three weeks before the month in question

In addition to running spike and slab regression to determine attribute importance, I also ran a gradient boosted machine (R package GBM). as of now, the model seems to be performing fairly satisfactorily, with the Mean Average Error = 5.4 (monthly sales are in the range of 10000), however there are clear seasonal trends in the difference between predicted variables and actual sales that seem to coincide with sales promotions and other promotional events. My question, therefore is:

How do I account for these seasonal trends while using GBM? or for a linear model?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.