基于协同算法的电子商务网站用户行为分析及服务推荐 毕业论文+答辩PPT+Python项目源码
[摘要] 互联网时代的到来和不断发展，逐步改变了人们传统的生活方式，社会也正朝着信息化、现代化的方向大步迈进。如今当人们遇到不懂的问题，首选已经不再是向老师或者身边的朋友咨询，而是向无所不知的百度询问。当人们需要购买东西时，也不用像之前一样只能去商场，在家里就可以利用电商平台轻松完成购买。人们的日常消费行为因为电商平台变得更为方便和快捷，但是大量信息的涌入也给用户带来一定的烦恼。用户如何快速在众多相似信息和商品中找到自己真正想要得到的内容，已经变得尤为重要。本文选取北京某家真实存在的法律网站为研究对象，进行用户行为分析，通过数据分析发掘出用户的浏览倾向，并对其进行个性化的网页推荐。本系统的核心算法为协同过滤推荐，并针对其冷启动问题，通过分析网站的实际情况和常见非个性化推荐算法的性能，最终确定了合适的非个性化推荐与协同过滤推荐相结合的推荐模型，极大地提高了推荐的准确率和推荐成功率。[关键词] 电子商务网站；用户行为分析；个性化推荐；协同过滤算法Abstract: The advent and continuous development of the Internet era have gradually changed people's traditional way of life, and the society is marching forward towards the direction of informatization and modernization. Nowadays, when people encounter problems that they don't understand, their first choice is no longer to consult their teachers or friends around them, but to ask the Baidu. When people need to buy something, they do not have to go to the shopping mall as before, but can easily finish at home through the e-commerce platform. People's Daily consumption behavior has become more convenient and fast because of the e-commerce platform, but the influx of a large amount of information also brings some troubles to users. How users can quickly find what they really want among the many similar information and products has become extremely important. In this paper, a real legal website in Beijing is selected as the research object to conduct user behavior analysis. Through data analysis, the browsing tendency of users is explored and personalized web page recommendation is made. The core algorithm of this system for collaborative filtering recommendation, and according to the cold start problem, through the analysis of the actual situation of site and common performance of the personalized recommendation algorithm, finally determine the appropriate personalized recommendation and the combination of collaborative filtering recommendation recommendation model, greatly improves the accuracy of recommendation and recommendation success ratio.Key words: E-commerce Website; User Behavior Analysis; Personalized Recommendation;Collaborative Filtering Algorithm目 录1 绪论 11.1 课题研究的背景与意义 11.1.1课题来源于背景 11.1.2 研究的目的和意义 11.2 国内外研究现状 11.2.1国外研究现状 21.2.2国内研究现状 22相关理论与关键技术 32.1个性化推荐系统概述 32.2协同过滤算法 32.2.1基于用户的协同过滤 32.2.2基于项目的协同过滤 32.2.3异同点分析 42.3物品相似度 42.3.1余弦相似度 42.3.2 Person相关系数 52.3.3 Jaccard相似系数 53 实验准备阶段 63.1系统需求分析 63.2可行性分析 63.2.1技术可行性 63.2.2经济可行性 63.2.3操作可行性 63.3数据集简介 63.4实验环境说明 63.4模型评价标准 74 系统具体实现 84.1数据探索分析 84.1.1网页类型分析 84.1.2点击次数分析 104.1.3热门网址统计 104.2数据预处理 104.2.1属性规约 104.2.2数据清洗 104.2.3数据变换 114.3模型构建 114.3.1数据集的划分 124.3.2测试集用户字典的构造 124.3.3用户物品矩阵的构建 124.3.4物品相似度矩阵的构建 124.3.5推荐列表的生成 134.4模型测评 135 模型测试与总结 14结束语 17致谢 18参考文献 19
