Recommender System: Recommendation Algorithms
- Non-Personalized Summary Statistics
- Mean-based:
- Background: In a 5-star scale rating system
- Symbols:
- Ui for users who rated item i
- rui for rating by user u to item i
- ∣Ui∣ for number of user who rated item i
- α for damping factor, larger would cause the rating floats more smooth
- μ for global rating across all items and users
- s(i) for score of item i
- Mean: s(i)=∣Ui∣∑u∈Uinrui
- Damped Mean: s(i)=∣Ui∣+α∑u∈Uinrui+αμ
- Association-based:
- Background: typically in a shop that customs buying products
- Symbols:
- P(i∣j): probability of buying i when already buying j
- Basic Association, measuring probability by counting:
- Story: How many percentage of people who buying i also buying j among the whole people who buying i?
- Formula: P(i∣j)=P(j)P(i∧j)=∣Uj∣/∣U∣∣Ui∩Uj∣/∣U∣
- Bad case: if j is popular, then the result is bad
- Bayes's Law: P(i∣j)=P(j)P(j∣i)P(i)
- Lift Association, measuring score by counting:
- Story: people who bought i and j together more often means i and j are more associative
- Formula: s(i∣j)=P(i)P(j)P(j∧i)
- Content-Based Filtering
- Information Filtering
- Knowledge-Based
- Collaborative Filtering
- User-User
- Item-Item
- Dimensionality Reduction
- Others
- Critique / Interview Based Recommendations
- Hybrid Techniques