With the explosive growth of network information,recommendation systems play a key role in alleviating the problems of information overload and information wandering.How to better uti-lize massive network information to mine user preferences and item characteristics has become a hot is-sue in current academic research.Targeting at this hot issue,the author put forward a context-aware collaborative filtering recommendation model based on circular convolutional neural network.The model uses recurrent convolutional neural network to mine item features in text auxiliary information,and combines probability matrix decomposition model to realize rating prediction.At the same time,it explores using multi-head attention mechanism to focus on multiple important information in auxiliary information.The model was tested on two publicly available datasets,ML-100k and ML-10m,and the experimental results showed that the proposed model had significant improvements in RMSE and MAE evaluation metrics compared to the widely used baseline model.Among them,the RMSE metric was 5.42%more effective on the ML-100k dataset than aSDAE which was the currently best.
维普
万方数据