Robust Recommendation Algorithm Based on User Suspicious Probability and Item Weight

With the extensive development of recommendation technology, the threat of shilling attacks faced by the existing collaborative recommendation algorithms is also increasing sharply. To face more and more complex shilling attacks, this paper constructs a robust recommendation algorithm that can resist shilling attacks from the perspective of recommendation algorithm. Existing robust recommendation algorithms usually improve robustness by sacrificing some recommendation accuracy and reduce the recommendation accuracy. To solve this problem, this paper proposes a robust recommendation algorithm based on user suspicious probability and item weight. Firstly, we establish the relevance vector machine classifier according to user profile features to evaluate user suspicious probability in the database. Secondly, we construct singular value decomposition algorithm based on Hebbian learning and matrix factorization algorithm by integrating user suspicion information. Finally, a dynamic weighting scheme is used in combination with the above algorithm and the collaborative filtering algorithm based on item weight, and the above algorithms are mixed according to a certain weight to obtain a robust collaborative filtering algorithm SRICF. By adjusting the weight ratio, advantages of each algorithm are brought into play, thereby improving recommendation accuracy and robustness of the algorithm.