Product Recommendations via Hybrid Recommender
This experiment trains a hybrid recommender to provide product suggestion to online retail customers based on past purchases/clickthroughs.
This experiment demonstrates how historical events like past purchases and ad clickthroughs can be used to generate pseudo-ratings suitable for training a collaborative filtering recommender system. A recommender is trained using the extracted "rating" feature, as well as descriptive features for users and products that permit content-based filtering. The trained recommender is then used to predict pseudo-ratings in a withheld test set.
