Research on Personalized Recommendation System Based on Machine Learning

Aiming at the problems of evolutionary learning of user preferences and high-dimensional sparse data processing in personalized recommendation systems. Inspired by the structural features of Hidden Markov Model, a personalized recommendation algorithm considering contextaware two-stage user preference set reasoning strategy) is proposed. By processing the historical scoring information of the system, the extraction process of user preferences is abstracted into a hidden Markov model to perform the first stage of user preference set learning and reasoning. Then, the occurrence probability of these scoring objects is weighted with the traditional item similarity calculation method to obtain the new similarity, and finally the recommendation result is generated. In the simulation experiment, the important parameters of the algorithm are trained, and compared with other algorithms, it is proved that the improved algorithm is effective.