【目的/意义】在大数据环境下,借鉴迁移学习概念,对移动图书馆收集的用户评分数据进行分析解决个性化推荐问题。【方法/过程】构建跨域协同过滤模型,共享跨域特征信息,重构评分矩阵实现个性化推荐。【结果/结论】案例结果表明跨域协同过滤模型能够解决移动图书馆推荐系统初期数据量较少的冷启动问题,并具有良好的应用性。同时实现了对用户多个领域的兴趣挖掘,增加了用户粘度。 [Purpose / Significance] In the context of big data, draw on the concept of relocation learning, analyze the user rating data collected by the mobile library to solve the problem of personalized recommendation. [Method / Procedure] Construct cross-domain collaborative filtering model, share cross-domain feature information and reconstruct scoring matrix to realize personalized recommendation. [Results / Conclusion] The results of the case show that cross-domain collaborative filtering model can solve the problem of cold starting with less initial data in mobile library recommendation system and has good applicability. At the same time to achieve the user a number of areas of interest mining, an increase of user viscosity.