基于增量学习和阈值优化的自适应信息过滤研究

为了适应实时在线的网络信息过滤需求，提出了一种新的自适应过滤模型。在系统的初始化阶段，运用增量学习方法对附加的少量伪相关文档进行学习，采用改进的文档词频方法来抽取特征词，以此扩展需求模板，提高模板准确度。在系统测试阶段，以系统效能指标最优为目标，提出了将概率模型和文档正例分布统计方法相结合来实现阈值优化的新算法。

Research on adaptive information filtering based on incremental learning and threshold optimization

In order to meet web-based on-line information filtering requirement, a new realistic adaptive information filtering model was proposed in this paper. In the system initiations stage, the profile is improved by active topic term learning and term weight value updating based on incremental learning the few pseudo relevant feedback samples. An incremental feature selection method was presented based on document frequency. In the filtering test stage, aimed at the system's optimization utility, a new technique was proposed which combined the probabilistic distribution model and document stream statistical information to update and explore the dissemination threshold actively. Experimental results show that the new methods lead to a higher performance in the adaptive information filtering system.