Goal of the project was to create a recommender system for a matchmaking website.
Project Details
As a final solution I have created a hybrid recommender system with 3 layers: rule-based, collaborative-based and content-based filtering. Rules were defined by client like male-female, limit for maximum age difference etc.. Then there was collaborative-based filtering which in this context was suggesting pairs together based on their previous choices. As last stage content-based filtering layer was defined as ML model. Ensemble model was created which consisted of Logistic Regression, Random Foresta and Neural Network.
Results were decent and matched to pairs that experts from company manual had been recommending.