ItemKNN
For this approach item vectors are represented as user-frequency [num users]-dimensional vectors constructed from the user-item interaction matrix . indicates how many times user interacted with the item .
Similarity Measurement
We employ cosine similarity to measure similarity between vectors:
where:
[num user num items]: original user-item interaction matrix
[num users num users]: user embedding matrix, where each row is:
- : set of -th user's (interaction timestamp, item index) pairs
- : last -th user's interaction timestamp
- : time decay coefficient (per second)
For instance, when is equal to 0 (which is the simplest case), Top-k recommendations are generated by retrieving the most closest (k) item embeddings to the latest user's representation.
Hyperparameters
Parameter hour defines the time period (in hours) associated with a decay factor of 0.9.
Implementation Note
For 5b-scale datasets, memory constraints arise when multiplying user-item interaction matrix and the user embedding matrix making computations infeasible. Thus, results for these datasets are not provided.