ItemCF
A branch of collaborative filtering that computes similarity based on co-occurrence relationships between items and uses it for recommendation. It is suited to scenarios such as e-commerce where the item set is stable and interpretability is required.
Timeline (last 90 days)
UserCF recommends based on user similarity, while ItemCF recommends based on item similarity; the two are different viewing directions (by row vs by column) of the same user-item interaction matrix.
ItemCF's co-occurrence matrix C[i][j] represents the number of users who have consumed both item i and item j; its size is the square of the number of items, but only non-zero entries need to be stored.
ItemCF's cosine similarity formula is sim(i,j) = C[i][j] / sqrt(N[i] * N[j]), and the denominator normalization naturally downweights popular items.
UserCF is suitable for content platforms where user interests diverge (news, video) and for trend-oriented scenarios; ItemCF is suitable for e-commerce where the item set is relatively stable and the number of users far exceeds the number of items, and for scenarios that require explainability.
Most e-commerce companies will prioritize ItemCF because the matrix is small, easy to maintain and naturally explainable.
ItemCF's explainability (directly showing "because you bought X") usually outperforms unexplainable recommendations on click-through rate.
All Facts (6)
UserCF recommends based on user similarity, while ItemCF recommends based on item similarity; the two are different viewing directions (by row vs by column) of the same user-item interaction matrix.
50%UnverifiedItemCF's co-occurrence matrix C[i][j] represents the number of users who have consumed both item i and item j; its size is the square of the number of items, but only non-zero entries need to be stored.
50%UnverifiedItemCF's cosine similarity formula is sim(i,j) = C[i][j] / sqrt(N[i] * N[j]), and the denominator normalization naturally downweights popular items.
50%UnverifiedUserCF is suitable for content platforms where user interests diverge (news, video) and for trend-oriented scenarios; ItemCF is suitable for e-commerce where the item set is relatively stable and the number of users far exceeds the number of items, and for scenarios that require explainability.
50%UnverifiedMost e-commerce companies will prioritize ItemCF because the matrix is small, easy to maintain and naturally explainable.
50%UnverifiedItemCF's explainability (directly showing "because you bought X") usually outperforms unexplainable recommendations on click-through rate.
50%