[Retail Notebook] Have a ML algorithm to classify new products

[SamTaz]

Notebook

https://github.com/atoti/notebooks/blob/master/retail/pricing-simulations-around-product-classes/main.ipynb

Enhancement

Current situation: A classification algorithm scores products and put them into 3 classes based on those scores. This classification algorithm runs on a set of receipts. Then KPIs calculation & pricing is performed

Improvement & context: Many retailers have about 10% of their catalogue renewed each year, meaning new products without receipts. If we train a ML algorithm that assigns a class to a product based on its characteristics (price, size, weight, brand, ...) on the already classified products, we can cover the new products issue. Idea would be somewhere in the notebook to have 10 new products coming, classify them thanks to the ML algorithm, and then price them depending on their class.

[jbe456]

A first step would be to introduce k-means clustering to generate the classification or a supervised learning algorithm for better accuracy.