多源数据融合的用户画像识别与推荐实证研究

Empirical Study on User Profile Identification and Recommendation with Multi-source Data Fusion

[目的/意义]构建多源数据融合获得细粒度的用户画像标签,加强和提升服务场景的用户画像识别与精准推荐,对于洞察用户需求、开展群体精准营销、提高用户忠诚度等有着重要的参考意义。[方法/过程]融入服务场景构建多源数据融合的用户画像识别与推荐分析框架,以用户价值模型RFM重构表征用户行为的RFCLS标签,采用LDA模型提取用户资源使用偏好的文本语义标签,继而将用户属性、用户行为和资源使用偏好等多源异构数据汇聚成用户画像标签体系后,选取随机森林模型对不同标签组合展开用户画像识别的模型训练和分类性能评估。[结果/结论]实证研究表明,与单一的数值型或者文本型画像标签体系相比,多源数据融合的用户画像模型提升了用户画像识别和分类的精准度,有效支撑高校图书馆开展更为针对性的营销服务策略和个体精准化服务推荐。

[Purpose/significance]Constructing a fine-grained user profile label through multi-source data fusion strengthens and enhances the identification and precise recommendation of user profiles in service scenarios.This has significant reference value for understanding user needs,conducting targeted group marketing,and increasing user loyalty.[Method/process]Integrating service scenarios,a framework for user profile identification and recommendation analysis based on multi-source data fusion is constructed.The RFCLS label,which redefines user behavior using the user value model RFM,is developed.The LDA model is employed to extract textual semantic labels that represent users’resource usage preferences.After converging multi-source heterogeneous data such as user attributes,behaviors,and resource usage preferences into a user profile label system,the Random Forest model is selected for model training and performance assessment in user profile identification with different label combinations.[Result/conclusion]Empirical research shows that compared to a single numeric or textual profile label system,the multi-source data fusion user profile model enhances the accuracy of user profile identification and classification.This effectively supports academic libraries in implementing more targeted marketing strategies and personalized service recommendations.