A Biased Proportional-Integral-Derivative-Incorporated Latent Factor Analysis Model

Nowadays, as the number of items is increasing and that users have access to limited, user-item preference matrices in recommendation systems are always sparse. This leads a data sparsity problem. The latent factor analysis (LFA) model has been proposed solution As basis LFA model, singular value decomposition (SVD) especially biased SVD great effects high-dimensional sparse (HiDs) matrices. However, it disadvantage requiring several iterations before convergence. Besides, PID-incorporated SGD-based (PSL) introduces principle discrete PID controller into stochastic gradient descent (SGD), learning algorithm model. It could solve problem slow convergence speed, but its accuracy needs be improved. In order make better solution, this paper fuses PSL with hoping obtain result by combining their advantages reconciling disadvantages. experiments show performs than traditional matrix factorization algorithms on different sizes datasets.