| 中国大学生的网络使用:基于大规模日志分析的模式识别新方法 |
| |
| 引用本文： | 严承希,王军,王珂.中国大学生的网络使用:基于大规模日志分析的模式识别新方法[J].图书情报工作,2019,63(14):83-93. |
| |
| 作者姓名： | 严承希 王军 王珂 |
| |
| 作者单位： | 北京大学信息管理系 北京 100871 |
| |
| 摘 要： | 目的/意义]深入挖掘和准确理解中国大学生日常网络行为模式，不仅对促进用户行为和检索领域的发展具有巨大的理论意义，而且在提升面向大学生用户的企业个性化服务与信息推荐能力方面也具有潜在的社会价值和实践意义。方法/过程]提出一种基于大规模日志分析的大学生用户行为模式识别新方法，该方法包括一种基于深度学习和文本分析技术的半监督学习算法"MaxMatching"以及混合两种特征熵（香农熵与真实熵）的聚类模型。结果/结论]实证结果表明本方法不仅在算法和结果解释上具有一定的优势，而且能从网络使用能力、访问时序性和主题倾向性三方面归纳与呈现中国大学生网络行为全方位模式。该方法和结论有效地拓展了信息检索领域查询项的语义化理解方面的方法体系，也为企业提升面向大学生用户的个性化信息推荐服务提供一定的参考和可行性意见。
|
| 关 键 词： | 中国大学生 网络行为 模式识别 大规模日志分析 |
| 收稿时间： | 2018-12-06 |
| 修稿时间： | 2019-03-06 |
Chinese College Students' Internet Use: A New Method of Behavior Pattern Recognition with Massive Log Analysis |
| |
| Yan Chengxi,Wang Jun,Wang Ke.Chinese College Students' Internet Use: A New Method of Behavior Pattern Recognition with Massive Log Analysis[J].Library and Information Service,2019,63(14):83-93. |
| |
| Authors: | Yan Chengxi Wang Jun Wang Ke |
| |
| Institution: | Department of Information Management, Peking University, Beijing 100871 |
| |
| Abstract: | Purpose/significance] It is of great significance to analyze and understand users' daily Web behavior patterns, which not only makes progress in the domain of user behavior analyse and information retrieval theoretically, but also has potential social values and practical significance in promoting personalized service and information recommendation for the undergraduate-oriented enterprises.Method/process] In this paper, a new method for college students' behavior Web pattern recognition based on large-scale log analysis was proposed. It included a semi-supervised learning algorithm "MaxMatching" based on deep learning and text analysis, and a hybrid model combined with two characteristic entropy (Shannon Entropy and Real Entropy).Result/conclusion] The empirical results showed that this method has the excellent performance in the algorithm and the result interpretation. Also, it can generalize and present all-round Chinese college students' Web behavior pattern in three aspects of network ability, temporality and topicality. The method and conclusion can effectively expand the methods about semantic understanding of queries in information retrieval, and provide some reference and feasible suggestions to undergraduate-oriented enterprises on personalized recommendation service. |
| |
| Keywords: | Chinese students online behavior pattern recognition massive log analysis |
|
| 点击此处可从《图书情报工作》浏览原始摘要信息 |
| 点击此处可从《图书情报工作》下载免费的PDF全文 |
