Collaborative Filtering
-
Cosine Based Similarity - to check similarity
-
Correlation Based Similarity - to check similarity
-
Distance Based Similarity - Euclidian or Manhattan Distance.
-
What to recommed - Create list of products that have not been bought and rank them based on ratings or highest rated or other popularity based criteria.
-
Negatives of Collaborative Filtering - 1. Memory based/ Lazy learning (Huge data matrix to be maintained also recommendation will be done only when cx vists website and then compute recommendation also send them to cx online/ offline), 2. Computation-Intensive (n^2 similarities).