In recent years,autoencoders have been widely used in the recommendation domain due to their strong data compression capabilities.However,research has shown that due to the sparsity of data in recommendation systems,autoen-coder models can develop biases during training caused by the lack of interactions between users and items,negatively im-pacting recommendation results.To address this issue,we propose a recommendation model based on item interaction con-straints.The model uses item interaction as a constraint and designs new parameter update rules to avoid biases introduced by data sparsity in model training.Additionally,the model incorporates item tag information into the training process,lever-aging new data sources to mitigate the effects of data sparsity and improve recommendation performance.Experiments on three datasets with varying sizes and sparsity levels demonstrate that the model is well-suited for sparse datasets,effectively enhancing recommendation accuracy and showing great application potential.
维普
万方数据