Martin, WJ; (2017) App Store Analysis for Software Engineering. Doctoral thesis , UCL (University College London).

Preview

Text App_Store_Analysis_Thesis_William_Martin.pdf Download (2MB) | Preview

Abstract

App Store Analysis concerns the mining of data from apps, made possible through app stores. This thesis extracts publicly available data from app stores, in order to detect and analyse relationships between technical attributes, such as software features, and non-technical attributes, such as rating and popularity information. The thesis identifies the App Sampling Problem, its effects and a methodology to ameliorate the problem. The App Sampling Problem is a fundamental sampling issue concerned with mining app stores, caused by the rather limited ‘most-popular-only’ ranked app discovery present in mobile app stores. This thesis provides novel techniques for the analysis of technical and non-technical data from app stores. Topic modelling is used as a feature extraction technique, which is shown to produce the same results as n-gram feature extraction, that also enables linking technical features from app descriptions with those in user reviews. Causal impact analysis is applied to app store performance data, leading to the identification of properties of statistically significant releases, and developer-controlled properties which could increase a release’s chance for causal significance. This thesis introduces the Causal Impact Release Analysis tool, CIRA, for performing causal impact analysis on app store data, which makes the aforementioned research possible; combined with the earlier feature extraction technique, this enables the identification of the claimed software features that may have led to significant positive and negative changes after a release.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item