【作者】 刘伟；

【导师】 郑世珏；

【作者基本信息】 华中师范大学 ， 计算机系统结构， 2016， 硕士

【摘要】 近年来,随着智能终端和移动互联网的迅速发展,移动端的社交应用每天都在不断涌现。微信由腾讯公司于2011年推出,目前微信用户已经突破6.5亿,覆盖国家超过200个,覆盖语言超过20种,微信已经成为移动端名副其实的巨无霸应用。微信公众平台作为微信的重要功能,目前已经吸引了数以千万计用户的入驻,微信公众平台提供了人们获取各类信息的重要渠道。本文主要通过分析微信公众号发布的热门内容与微信公众号自身的名称特征及介绍信息,设计微信公众号内容推荐系统,对微信公众号给出较为精确的推荐结果,以便公众号能够经常发布高质量的图文内容,吸引用户的持续关注。本文使用基于内容和协同过滤的混合推荐算法,对微信公众号进行精确的内容推荐,使微信公众号能够经常发布高质量的内容,不断获取用户流量。对微信公众号的推荐主要从两个方面进行：一方面,使用基于内容的推荐算法,利用爬虫技术获取指定微信公众号发布的所有内容,通过分析该微信公众号发布文章中的所有热门内容,提取热门文章的关键词并进行词频统计,将微信公众号文本内容向量化,使用文本相似度算法从文章库中选取与该微信公众号热门文章主题相似度较高的文章进行推荐；另一方面,使用协同过滤的推荐算法,通过分析微信公众号自身的名称特征及介绍信息,从微信公众号数据库中提取出与指定微信公众号特征相似度较高的公众号,将该有较高相似度公众号的热门内容推荐给指定微信公众号,完成协同过滤的推荐过程。基于内容的推荐算法与协同过滤推荐算法都是经典的推荐算法,但是单一的推荐算法在实际应用中都有一定的缺陷。在实际应用中,通常使用的推荐算法是混合推荐算法,即将多种推荐算法同时应用,从而得到更为精确的推荐结果。本课题使用的推荐算法即为基于内容推荐和协同过滤推荐的混合推荐算法,并对其中的协同过滤推荐算法进行了一定的改进。通过实验仿真,得出本文设计的微信公众平台推荐系统的最终推荐结果,分析推荐系统得到的推荐结果与微信公众号热门文章的相似度,证明最终的推荐结果满足推荐要求,并且证明本文改进的混合推荐算法在微信公众平台得到的推荐结果符合预期。更多还原

【Abstract】 In recent years, with the rapid development of smart terminals and mobile Intemet,social applications on mobile terminal emerge in large numbers every day. WeChat was rolled out by Tencent in 2011, and the number of WeChat users has exceeded 650 million this year. WeChat supports more than 20 kinds of language and has been used over 200 countries, thus becoming a veritable super application on mobile terminal. As a significant part of WeChat, WeChat Public Platform changes the way people getting information. This article mainly analyses hot con-tent, self features and introduction information of WeChat Official Account, design a content recommendation system for WeChat Official Account,Which can provide highly-related and up-to-date information for a given WeChat Official Account, thus assisting the WeChat Official Account in publishing high-level content and attract more users’subscription.This article use hybrid recommendation algorithm based on content-based recommendation algorithm and collaborative filtering recommendation algorithm to provide WeChat Official Account precise results, which can assist WeChat Official Account in gaining user’s traffic. The core of this recommendation system is mainly based on two parts. On the one hand, the system use content-based recommendation algorithm and use crawler to get all contents related to the appointed WeChat Official Account. Then we extract key words of all hot articles, count the key words frequency and vectorize WeChat Official Account text con-tent. On the other hand, the system use collaborative filtering recommendation algorithm to analyse the hot contents, self features and introduction information of a given WeChat Official Account, and extract other WeChat Official Account that are similar with the appointed WeChat Official Account from database. Then the system provides the hot contents from the extracted WeChat Official Account to the appointed Official Account.The content-based recommendation algorithm and collaborative filtering recommendation algorithm are all classical recommendation algorithm, but each single recommendation algorithm has its own shortcoming. So we often use hybrid recommendation algorithms in practical applications, in other words,we usually use more than two recommendation algorithms in practical appli-cations. This article use the hybrid recommendation algorithm based on content-based recommendation algorithm and collaborative filtering recommendation algo-rithm, and improve the collaborative filtering recommendation algorithm.By comparing the similarity between recommended results and hot contents, the simulation experiments shows that the recommendation system meets the requirements.In addition, the experiment proves that the unproved algorithm works well in this recommendation system.更多还原