MKR模型在电影推荐系统中的应用研究

Research on the Application of MKR Model in Movie Recommendation System

针对电影推荐系统中推荐结果的准确性和可解释性不高的问题,研究了基于多任务特征学习的知识图谱增强推荐(multi-task learning for knowledge graph enhanced recommendation, MKR)。通过构建知识图谱,将其作为辅助信息构建了MKR模型,并将其应用到电影推荐系统中。采用预测用户满意度评分的方法根据评分结果来判定用户对电影的喜好程度,并将合适的电影类型推荐给用户。最后将MKR模型与几种常见的推荐模型进行比较,使用不同的评价指标进行预测,并在top-K场景中比较了不同K值下各个模型的推荐性能。实验结果表明,MKR模型在电影推荐系统中有良好的表现,在准确率和推荐结果上均优于其他模型,显著提升了推荐的性能。

Aiming at the low accuracy and interpretability of recommendation results in movie recommen-dation system, a multi task learning for knowledge graph enhanced recommendation (MKR) based on multi task feature learning is studied. By constructing a knowledge graph, and using it as auxil-iary information to construct a MKR model, and applied to the movie recommendation system. Using the method of predicting user satisfaction score, according to the score results to determine the user’s preference for movies, and recommend the appropriate movie types to users. Finally, MKR model is compared with several common recommendation models, and different evaluation indexes are used for prediction. The recommendation performance of each model under different K values is compared in the Top-k scenario. The experimental results show that MKR model has a good performance in the movie recommendation system, and is superior to other models in accu-racy and recommendation results, which significantly improves the performance of the recom-mendation.