utiliser le Collaborative Filtering pour proposer des paniers

[evaletolab]

Note:

Collaborative Filtering is a generic approach that can be summarized as "using information from similar users or items to predict affinity to a given item". There are many techniques that can be used for Collaborative Filtering. The two that are most well-known and discussed in the literature are Nearest Neighbors (knn) and Matrix Factorization (MF). Knn is clearly a supervised method. As for MF, depending on the details of its usage one can call it supervised, unsupervised, or semi-supervised.

Context

• user orders
• user likes
• products update (new product in a category that match user interest should be suggested)
• products promotion (Promotional products that match user interest should be suggested )

Objectives

The main goal of this application is to make user navigation more friendly. Selection of products should be automated, it should simply match the user interest

• user to products (or category/products)
• user to categories
• product to products

Conclusion

• Compute Collaborative filtering (user=>items,item=>items) (with or without category)
• Workflow for products update time
• Workflow for products promotion
• Workflow for products image detection (image=> tags)

Literature

• https://www.analyticsvidhya.com/blog/2018/06/comprehensive-guide-recommendation-engine-python/

Resources

• https://www.npmjs.com/search?q=recommendation%20engine
• https://github.com/Recombee/node-api-client
• https://github.com/iceonepiece/jonfon
• https://github.com/proin/nodeml
• https://www.npmjs.com/package/product-recommender