One of the important retailer’s machine learning use cases is sales forecasting. This in fact is one of the first machine learning pipelines that VirtusLab has created for Tesco. The first approach a data scientist would think of, when trying to sketch the simplest solution, would probably be some model from their statistical toolbox (perhaps based on moving average).
But how would a neural network cope with such a forecast? How would you go about wrangling the data, setting up a local dev environment, scaling your training? Is it worth it? These and some other questions I will attempt to answer during this presentation.
[Optional] As this will be a hands-on presentation, showing working models, you will be able to code along if you wish. Here is the link with setup instructions: https://github.com/gregaw/sales-forecasting-with-nn
Grzegorz is the Head of Data Science at VirtusLab where he's focusing on putting data to good use.
He's mostly interested in data, algorithms, software engineering, distributed systems, machine learning and quantitative finance.