Recommender systems · Module 3

Building Your
Recommendation Model.

Get hands-on: implement a real recommendation engine with collaborative filtering, matrix factorisation (SVD) or neural networks. From theory to predictions in production.

Python

scikit-surprise
+ TensorFlow

0^.84

Estimated RMSE
MovieLens

Hybrid

Multi-algorithm
Ensembling

— Simulator

Configure your architecture

Interactive

User-based collaborative filtering

User-based CF · Cosine similarity

Item-based collaborative filtering

Item-to-item · Optimal scalability

SVD / NMF factorisation

Matrix factorization · Latent factors

Neural networks (Neural CF)

Deep learning · Non-linearities

Selected configuration:

— Estimated performance

Hybrid approach enabled · Estimated accuracy: 0.84–0.92 RMSE on a MovieLens-like dataset

Example of a generated recommendation: "Marie liked Inception, suggested: Interstellar + Dune: Part Two (score 4.6/5)"

Data Collection & Preparation

Computing & Generating Recommendations

Building YourRecommendation Model.

Configure your architecture

Interactive model configurator

Building Your
Recommendation Model.