A Novel Item Recommendation Framework based on Deep Learning Algorithm
Chunrong Yao · 2023
With their unique adaptive characteristics, neural networks can effectively solve unstructured problems such as patterns and complex information. They have been widely used in neural expert systems, convex optimization and prediction, and other fields. The organic integration of neural networks and other conventional research methods will greatly promote the development of artificial intelligence, information analysis and other fields. In this study the deep neural network is applied to the task of item recommendation. Efficient information search is the initial step for the task of item recommendation, the Top-K sorting algorithm is constructed. Top-K sorting is based on the List-wise method, the List-wise method uses all document lists as the samples of the input space, also called the document list method, and its output is a sorted list. This can be treated as the pre-processing of the data. Then, the graph neural network is selected as the deep structure, it is a general term for learning methods and also the core idea is to continuously update the representation of the nodes on the basis of the message-passing mechanism. In the experiment, MovieLens-20M, MovieLens-1M and the BookCrossing data sets are used to verify the proposed algorithm. Recall and AUC are used as the performance indicator, and compared with the state-of-the-art algorithms, the proposed model outperformed.