This repo demonstrates a useful pattern for building many closely-related machine learning models efficiently.

see my blog post

We are Acme, Inc., a company that helps chain retailers forecast sales. This repo is our framework for generating predictive models based on each of our clients' historical data.

We assume that Acme has a roughly-standard data format, and asks clients to supply data in that format. Here, we simulate that process by pulling data from Kaggle and transforming it into the format that a client's application may natively use -- a hierarchical structure similar to typical normalized relational data or object structure.