基于潜在因子模型的时间序列语义信息挖掘及匹配方法研究

A Personalization Recommendation Method with Temporal Context

【目的/意义】当前全球信息化时代下信息过载问题日趋严峻,在深度挖掘信息的基础上,结合用户行为特征进行智能匹配显得尤为重要。【方法/过程】本文在基于潜在因子模型的个性化推荐算法的基础上,构建了结合时间序列的语义信息挖掘及匹配模型。通过引入用户历史行为的时间序列语义信息,提高已有模型预测用户偏好的准确性,结合因子分解机的思想实现对扩展模型的构建,并通过movielens数据集对该方法的有效性进行验证。【结果/结论】实验结果表明,新模型能够有效提高已有推荐模型预测用户偏好的准确性,从而实现了良好的数据挖掘及匹配效果。

【Purpose/significance】Personalization recommendation can effectively solve the negative influence of information overload to users in the environment of big data. And the existing personalization recommendation model is insufficient in integrating the temporal context of users＇ behavior.【Method/process】We build a new extend model of personalization recommendation based on the method of latent factor model. Temporal context of users＇ historical behavior is introduced into new model to improve the predictions＇ accuracy. We implement the new model based on the method of factorization machines,and verify the validity of new model by using the data of movielens dataset.【Result/conclusion】Experimental results demonstrate better performance of new model in improving the predictions＇ accuracy of users＇ preferences compared with existing model.