Unified probabilistic matrix factorization of follow

We add the timeliness of interaction behavior into the traditional probability matrix factorization algorithm to establish the newly proposed user influence model Combined with the static attention relationship, a recommendation algorithm based on unified probabilistic matrix factorization is proposed. The experimental results on the data set established by Sina Weibo show that compared with NMF, PMF, SoRec and other algorithms, the proposed algorithm has an average increase of 11.82% on the F1-Measure index under different data densities.