Recommendation System
Content-Based Filtering
Approaches
Classification
Regression
Collaborative Filtering
Approaches
Model-based
Matrix Factorisation
Memory-based
user-user
item-item
Problems
Cold Start Problem
Representative based
use a subset of items and users that represent the population.
Content-based
use side information such as text, social networks, etc.
Bandit
consider the exploration vs exploitation tradeoff in new items.
Deep Learning
recent methods that try to solve some of the issues tackled above but using a black box.
Solution - Content + Collaborative = Hybrid Filtering
Grey Sheep Problem
Solution - Content + Collaborative = Hybrid Filtering
Evaluation
Metrics-based Evaluation
Human-based Evaluation
Last updated