Support attack detection algorithm for recommendation system based on deep learning.

In order to enhance the performance of recommendation systems that support attack detection algorithms, we have designed a novel approach based on deep learning. Specifically, our algorithm focuses on improving convergence, user scoring accuracy, algorithm efficiency, and detection accuracy. To achieve this, we first construct a preliminary user rating matrix, which is optimised by incorporating the user preference word matrix and the weight of the preference word. Additionally, we adjust the size of the matrix using principal component analysis. Next, we construct a deep, bidirectional RNN model based on the deep learning network. This model is then combined with the user scoring matrix to identify the type of user. In the case of abnormal or false users, our algorithm is able to identify the recommendation system's support attack through the detection results. The experimental results demonstrate the effectiveness of our algorithm. Specifically, our approach achieves fast convergence speeds, with the loss value remaining low throughout the process. Moreover, we achieve high average accuracy in user scoring, with a score of 97.14%. The detection time of the recommendation system support attack is also consistently lower than 0.7 s. Furthermore, our approach achieves an average accuracy of 98.09% in the detection of recommendation system support attacks. Overall, our algorithm shows promising results for practical applications in the field of recommendation systems. [ABSTRACT FROM AUTHOR]