Restricted Boltzmann Machine Collaborative Filtering Recommendation Algorithm based on Project Tag Improvement

摘要

The collaborative filtering algorithm based on Restricted Boltzmann Machine (RBM) has the problem of heavy weight in the prediction of the "popular project" and poor discrimination of the "unpopular project", which results in reduced prediction accuracy of the model algorithm. In order to improve the personalization and accuracy of the model, this article integrates project tags into the prediction process based on the RBM model and uses the project tags to describe the user's own interest preference, which strengthens the individual needs of the user: First, it uses projects that the user has already graded to calculate the user's probability of rating the objective tag; Second, it uses the probability of the scoring to predict the probability of different scoring levels of the user's unprotected items; Then, RBM model training is used to predict the probability that the user will score different grades for items that are not scored; Finally, the two scoring probabilities are weighted to the RBM model prediction process to produce prediction results. Experimental results using Movielens datasets show that the accuracy of the proposed method is improved by 1.2% compared with the original algorithm.