A form of collaborative filtering based on the similarity between items calculated using people’s ratings of those items
Why using Item CF rather than user CF
The number of users weighs more than number of products
Item will not change frequently, lowering calculation
Using user’s historical data, more convincing
How to implement Item CF
Build co-occurrence matrix
Build rating matrix
Matrix computation to get recommending result
How to define similarity between two movies
If one user rated two movies, these two are related
Co-Occurrence matrix:
A co-occurrence matrix is a matrix that is defined over an image to be the distribution of co-occurring pixel values (grayscale values, or colors) at a given offset.
