versesrev / MC-AGA

Low-Rank Matrix Completion over Finite Abelian Group Algebras for Context-Aware Recommendation

The paper is to appear in the Proceeding of ACM International Conference on Information and Knowledge Management (CIKM) 2017.

This repository contains:

1. Appendix.pdf: Proof of theorems in the paper.

2. MCAGA.m: A MATLAB implementation of MC-AGA, the algorithm proposed in the paper.

3. demo.m: A simple demonstration of MC-AGA on a random matrix.

4. Scripts that reproduce our results in the paper.

• For frappe dataset:

  • MCAGA_frappe.m: The main script.
  • metrics.m: The function for evaluation metrics.

• For sushi dataset:

  • MCAGA_sushi.m: The main script.
  • sushi_explicit2.m: The script for preprocessing data, including discretization and data splitting.
  • sushi_explicit3.mat: The data used by the main script.
  • rng_state.mat: The rng state used by the preprocessing script.

The frappe dataset should be downloaded from http://baltrunas.info/data/CARS2_code.zip and placed under the CARS2_code directory.

The sushi dataset should be downloaded from http://www.kamishima.net/sushi/sushi3-2016.zip and placed under the sushi3-2016 directory. If unavailable, try retrieving it from https://archive.org/web/