Collabrative Filtering Recommendation including Item-based CF and User-based CF, implementing with Sklearn and Surprise.
Similarity function: pairwaise consine distances
Item-based CF Prediction:
User-based CF Prediction:
Evaluation using Root Mean Square Error:
Similarity function: Pearson Baseline
Item-based CF Prediction: Recommending 10 items via 1 item based on Pearon Similarity. That 10 items are neighbors of the given item using KNN method
User-based CF Prediction: Recommending 10 items via 1 user based on Pearson Similarity. That 10 items are the top rating items of the given user's neighbors, whose count is also 10. In the program I simply make a item list of given users' neighbors and select top 10 rating items.