针对协同过滤推荐系统在稀疏数据集条件下推荐准确度低的问题,提出了推荐支持度模型以及用于该模型计算的邻域线性最小二乘拟合的推荐支持度评分算法(linear least squares fitting,LLSF)。该模型描述用户对被推荐项目更感兴趣的可能性,通过用高支持度的评分估计取代传统的期望估计法来找出用户更喜欢的项目,从而提高推荐的准确度,并从理论上论述了该算法在稀疏数据集条件下相对其他算法具有更强的抗干扰能力。该模型还易于与其他推荐模型融合,具有很好的可拓展性。实验结果表明:LLSF算法显著提升了推荐的准确性,在MovieLens数据集上,F1分数可达到传统的kNN算法的3倍多,对于越是稀疏的数据集,准确率提升幅度越大,在Book-Crossing数据集上,当稀疏度由91%增加到99%时,F1分数的改进由22%提高到125%。同时该方法不会牺牲推荐覆盖率,可以保证长尾项目的挖掘效果。
Most of the existing approaches focus on identifying mismatches and synthesizing adaptors at design-time or recently at run-time. However, few works have been proposed to support adaptor reconfiguration when services in the composition evolve due to changes in business needs. To address the deficiencies, the problem of adaptor reconfiguration is targeted in the context of service composition. Firstly, the formal models for describing services and adaptors are presented. Then, under this formalization,the notion of reconfiguration compliance is proposed to determine the validity of an adaptor instance with respect to its history executions and future executions. Based on the notion,the algorithm for reconfiguration analysis of adaptors is presented and it can be used for determining the migratability of an adaptor instance and the corresponding target state of reconfiguration if migratable.Finally,feasibility of the proposed approach is validated on a realistic case study. The proposed approach improves the flexibility of adaptor-based service composition by equipping adaptors with reconfiguration capabilities.
