A Method of Bayesian Probabilistic Matrix Factorization Based on Generalized Gaussian Distribution

Yan Cairong,Zhang Qinglong,Zhao Xue,Huang Yongfeng
DOI: https://doi.org/10.7544/issn1000-1239.2016.20160582
2016-01-01
Abstract:The method of Bayesian probability matrix factorization (Bayesian PMF) is widely used in the personalized recommendation systems due to its high prediction accuracy and excellent scalability . However ,the accuracy is affected greatly by the sparsity of the initial scoring matrix .A new Bayesian PM F method based on generalized Gaussian distribution called GBPM F is proposed in this paper .In the method ,the generalized Gaussian distribution (GGD) is adopted as the prior distribution model in w hich some related parameters are adjusted automatically through machine learning to achieve desired effect .Meanwhile ,we apply the Gibbs sampling algorithm to optimize the loss function .Considering the influence of the time difference of scoring in the prediction process ,a temporal factor is integrated into the sampling algorithm to optimize the method and improve its prediction accuracy . T he experimental results show that our methods GBPM F and GBPM F‐T can obtain higher accuracy w hen dealing with both sparse matrix and non‐sparse matrix ,and the latter can even get better effect .When the matrix is very sparse ,the accuracy of Bayesian PM F decreases sharply w hile our methods show stable performance .
What problem does this paper attempt to address?