Recommendation algorithm based on Explicit and Implicit feedback Matrix factorization

Matrix factorization is an effective solution to sparse data problem in the recommendation algorithm, but it relies too much on the user's direct behavior, preventing the algorithm from representing the user's true preference for an item. In order to address this challenge, this paper establishes an implicit feedback model through the movie view duration based on explicit feedback. The alternating least-squares method and its optimization method ALS-WR are introduced to factorize the explicit feedback model and the implicit feedback model respectively into two user-latent factor matrices and two item-latent factor matrices. Then the concept of weight-value is introduced to integrate each pair of user-latent factor matrices and item-latent factor matrices to finally produce recommendations. The simulation experiment on the MovieLens dataset is compared with the ALS-based, ALS-WR-based, and SVD-based recommendation algorithms. The experiment results show that the algorithm proposed in the paper is superior to the above algorithms in terms of mean absolute error (MAE) and root mean square error (RMSE), and is effective in improving the quality of recommendation.