多种搜索行为联合分析方法研究  

Joint Analysis Method of Multiple Search Behaviors

作　　者：张引[1,2] 高克宁[1] 张斌[1] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819 [2]东北大学医学影像计算教育部重点实验室,辽宁沈阳110819

出　　处：《东北大学学报（自然科学版）》2014年第9期1249-1252,共4页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(61073062);沈阳市科学技术计划项目(F11-264-1-35;F11-264-1-33);中央高校基本科研业务费专项资金资助项目(N120304002)

摘　　要：针对现有搜索行为分析方法只能分析单一类型搜索行为,无法有效获取用户兴趣的问题,提出多种搜索行为联合分析方法.通过使用M5模型对页面停留时间、鼠标点击次数、页面重访问次数及滑块移动次数4种类型的用户行为进行联合分析,从多个角度获取用户行为信息用于分析用户兴趣,并实现了对多种搜索行为构成的高维数据进行联合分析,同时满足了在线行为分析中对结果计算的实时性要求.实验表明该方法可以提供比Belkin方法更高的行为分析质量.Aiming at the fact that available search behavior analysis methods could only analyze one single type of search behavior, which leads to the problems that the user interest could not be effectively get, a joint analysis method of multiple search behaviors was proposed. By combining the analysis of page dwell time, mouse click times, page revisit times and slider move times using M5 model, clues from more angles other than one single type of user behavior were gained for user interest analysis, and joint analysis of high dimension data composed of multiple user behaviors was realized with timeliness ensured for online behavior analysis. Experiment results show that the proposed method can analyze the behaviors better than compared approaches.

关 键 词：搜索引擎 个性化搜索 用户行为分析 搜索行为特征 用户行为模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]