Creating Rossman type sales forecasting datasets

If we are trying to do sales forecasting for a store/biz like Rossman using neural nets, would we get a better model if we included advertising data (eg amount spent on ads per day) and traffic data (eg number of people visiting the store per day) while initially creating the dataset? Almost all businesses / stores advertise… And sales could also depend on number of ppl visiting the store daily… so wondering if those details should be included while making a ‘perfect’ dataset?

1 Like

Advertising: yes.

People visiting: not sure. You can’t use it as a predictor (since you don’t know it ahead of time). But it could perhaps be a second target variable.

1 Like

Thanks. Makes sense. Just curious, do you have a rough % split on the amount of time you would spend understanding the dataset as compared to training/building a model? In a real world project…

The two go hand in hand, and can’t be separated. See my machine learning videos for many hours of discussion on this… :slight_smile: