移动App的用户需求与版本变迁的潜在关系挖掘与分析

发布时间：2018-07-09 21:57

本文选题：移动App + 应用商店　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：随着互联网的兴起和智能手机的流行,市场上已经有数以百万计的移动App可以安装在用户手机上并为其提供服务。用户通过应用商店下载、使用这些App,同时在应用商店中以评论的方式对App的质量进行反馈。用户的评论对于App开发者至关重要。开发者对App进行更新时除了要考虑App本身的发展需求,更需要考虑用户的需求与感受。将高质量评论中所体现的信息融入到App更新中将有助于提高App的质量和评级。然而目前为止,尚未有明确研究表明开发者是否利用、以何种程度利用用户评论信息对自身App进行改进,从而提升服务质量。针对以上问题,本文通过提取用户评论和App更新日志中的特征并识别它们之间的潜在关系,利用对原子更新单元(Atomic Update Unit,abbr.AU)进行聚类的办法发现了7种更新模式(Update Pattern,abbr.UP)。更新模式体现了开发者以何种强度在何种及时程度以及充分程度上对用户请求进行响应的一种共性行为模式。同时,针对更新模式进行了一系列的实证研究。本文的结论帮助开发人员清楚地了解到自身对用户评论进行反馈的习惯,对开发人员如何充分利用用户评论提升App质量提供了建议。具体研究内容包括以下几个部分:(1)实现了一种针对Google Play应用商店上App的更新日志和评论的自动化收集工具。详细介绍了工具在云端的部署和维护的过程。定义了待收集App数据的数据模型。对收集到的数据进行了统计层面的分析并得到了一些统计结果。(2)定义了原子更新单元(Atomic Update Unit,abbr.AU)并介绍了其生成方法,给出了针对原子更新单元的及时性、充分性、特征更新强度、用户特征请求强度变化趋势的计算方法。设计了一种针对用户特征请求强度/特征更新强度变化趋势(Intensity Trend Chart for Feature Request and Update,abbr.TC)的分段拟合归一化算法。(3)定义了更新模式(Update Pattern,abbr.UP)。给出了挖掘更新模式的方法,挖掘得到了七种更新模式并对其进行分析。(4)进行了一系列实证研究,得到了一些结论:App开发者对某特征进行更新时采用哪种模式很大程度上取决于开发者本身喜好而不是该特征的本质;开发者采用更新的稳定性存在明显分化,约有65%的App的的更新稳定性处于较较不稳定状态,与此同时,有12%的App的更新稳定性处于非常稳定状态,这部分开发者更倾向于对自身App的同一特征在历史更新中采用相同的模式;发现了两种模式与App评论数量有显著正相关,发现三种更新模式与App在应用商店中的排名有负相关。发现了一种更新模式与App的评分有正相关。
[Abstract]:With the rise of the Internet and the popularity of smartphones, there are already millions of mobile App on the market can be installed on user phones and provide services. Users download it from the app store, use it, and comment on the quality of App in the app store. User comments are critical to App developers. Developers need to consider not only the development needs of App itself, but also the needs and feelings of users when updating App. Incorporating the information embodied in high-quality reviews into App updates will help improve the quality and rating of App. However, up to now, there is no clear research on whether or not developers use the user comment information to improve their App, so as to improve the quality of service. In order to solve the above problems, by extracting the features from user comments and App update logs and recognizing the potential relationships between them, seven update patterns (update abbr.up) are found by clustering Atomic Update unit abbr.AU. The update pattern reflects a common behavior pattern in which the developer responds to the user's request in what degree of timeliness and adequacy. At the same time, a series of empirical studies on the renewal model are carried out. The conclusion of this paper helps developers to understand their habit of feedback on user reviews and provides suggestions on how to make full use of user reviews to improve the quality of App. The main contents of this paper are as follows: (1) an automatic collection tool for App update logs and comments in the Google play App Store is implemented. The deployment and maintenance of the tool in the cloud are introduced in detail. The data model of App data to be collected is defined. The collected data are analyzed at the statistical level and some statistical results are obtained. (2) the Atomic Update unit (AU) is defined and its generating method is introduced, and the timeliness, adequacy and characteristic update intensity of the atomic update unit are given. The calculation method of intensity change trend of user feature request. A segmented fitting normalization algorithm for intensity trend Chart for feature request and Updateabbr.TC is designed. (3) Update pattern is defined. In this paper, the method of mining update patterns is given, and seven updating patterns are obtained and analyzed. (4) A series of empirical studies are carried out. Some conclusions are drawn: the pattern used by App developers to update a feature depends to a large extent on the nature of the feature rather than the nature of the feature; the stability of the developer's adoption of the update is clearly divided. About 65% of the App have relatively unstable renewal stability, while 12% of the App are in a very stable state. This part of developers tend to use the same pattern for the same characteristics of their own App in the historical update, and found that the two models have significant positive correlation with the number of App reviews, and the three update patterns have negative correlation with the ranking of App in the application store. A positive correlation was found between an update model and the App score.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.56

【参考文献】