Digital Library Book Recommendation Model Based on Collaborative Filtering and Cloud Computing

Abstract

With the rapid development of the Internet, digital resources are growing exponentially. In this context, more libraries are transforming into smart libraries. The demand for detailed project recommendations and real-time updates is becoming increasingly prominent due to information redundancy caused by excessive data. Therefore, this study attempts to build a digital library data platform based on cloud computing. In the traditional recommendation algorithm, the attention mechanism is introduced. A neural collaborative filtering algorithm based on channel attention is proposed, and an improved digital library book recommendation model is designed by combining the two. The test results showed that the average value, optimal value, and standard deviation of the improved algorithm were the lowest among the comparison algorithms. The loss function value was the lowest. The average recommendation accuracy of the model was 92.08%, the recall was 89.88%, and the area under the curve was 0.89. When the number of recommended books was 5, 10, 15, and 20, the recommendation matching degree was 92.06%, 96.27%, 90.03%, and 93.46%, respectively. The coverage rate was above 90%, with an average time consumption of 0.28s. The recommended running time on small, medium, and large data sets was 2.3s, 10.8s, and 21.7s, respectively, and the memory consumed was 119MB, 481MB, and 961MB, respectively. The mean reciprocal ranking for different proportions of experimental population was 0.393. It demonstrates that the proposed book recommendation model for digital libraries has higher accuracy, stable recommendation effects, and less time consumption. The library recommendation model can effectively provide targeted and personalized recommendation services to users.