Hybrid recommendation algorithm based on logistic regression refinement sorting model

摘要

The mainstream recommendation systems mainly use content-based method or collaborative filtering method. However, in specific recommendation scenarios, hybrid algorithm often performs better than single algorithm. We introduce a new recommendation method based on hybrid algorithm, which combined with logistic regression refinement sorting model. Our method can achieve higher accuracy rate and recall rate when we need to consider item and user features comprehensively. We recall and sort items by the hybrid algorithm based on content-based method and collaborative filtering method. After recalling process, we obtain preliminary rough sorting recommendation lists. Then we use logistic regression refinement sorting model to train the rough sorting results. The recommendation results can be more accurate after refinement sorting. We used the song data of a music website as experimental data and set three comparative experiments under different feature weight values. The experimental results show that when we consider the item and user features comprehensively, our method is better than other mainstream methods in accuracy rate and recall rate.