Research Of Depth Recommendation Based On Probabilistic Matrix Factorization

Probabilistic Matrix Factorization(PMF)is applied in the recommendation system and solves the problem of poor performance due to the traditional Matrix Factorization when the scoring is very sparse.While regularizing the traditional Matrix Factorization,PMF introduces a probabilistic model to optimize the recommendation results.However,in practice,both the user's preferences and the characteristics of movies change over time.Therefore,only use of the probability matrix decomposition model can not accurately capture the features of users and movies in practical applications,and cannot meet the requirements of the industry and academia.To solve the problem mentioned above,We propose in this paper a model based on PMF and Long Short-Term Memory(LSTM),which can accurately capture features of users and movies over time named PMF-lstm.First of this paper,it introduces the related theories of this parper,including the methods commonly used in the recommendation system;Matrix Factorization,Probabilistic Matrix Factorization,Deep learning;and the training optimization method of Deep Learning model.Then,based on PMF and LSTM,we propose the PMF-lstm model.Because LSTM is a time-recursive neural network that is suitable for processing and predicting important events with relatively long intervals and delays in time series.Therefore,the combination of PMF and LSTM can accurately capture the characteristics of users and movies that change over time,thereby improving the performance of the model in the recommendation system for the dynamic detection of users and movies.At last,through experiments,our model is compared with the existing methods,and the experimental results are analyzed to verify the feasibility of this method and the improvement of the accuracy.