[Recommendation System]--Collaborative Filtering Algorithm

Collaborative filtering

Collaborative filtering mainly includes user-based collaborative filtering and commodity-based collaborative filtering

1. User-based collaborative filtering

a. Find other users closest to the user, find items they have watched/bought but the current user has not watched/bought, and score according to the distance weighted.
b. Find the items with the highest score to recommend

2. Item-based collaborative filtering

According to the user's behavior on the product/content, the similarity between items is calculated, and the closest item to the current item is found for recommendation.

3. Similarity/Distance Metrics

• Euclidean distance $$D_{1,2}=\sqrt{(x_1-x_2)^2+(y_1-y_2)^2}$$
• Jaccard similarity $$J(A,B)=\frac{{|A \cap B}}{|A \cup B|}$$
  Narrow Jaccard similarity, calculate the similarity between two sets, the "value" of the element is 0 or 1
• cosine similarity $$cos(\theta)=\frac{a^Tb}{|a|·|b|}$$
• Pearson similarity