Android App Store (Google Play) Mining and Analysis

Abstract: The aim of mining and analysis of Apps in Google Play, the largest Android app store, is to provide in-depth insight on the hidden properties of the repository to app developers or app market contributors. This approach can help them to view the current circumstances of the market and make valuable decisions before releasing products. To perform this analysis, all available features (descriptions of the app, app developer information, app version, updating date, category, number of download, app size, user rating, number of participants in rating, price, user reviews and security policies) are collected for the repositoryand stored in structured prole for each app. This scientic study is mainly divided into two approaches: measuring pair-wise correlations between extracted features and clustering the dataset into number of groups with functionally similar apps. Two distinct datasets are exploited to perform the study, one of which is collected from Google Play (in 2012) and another one from Android Market, the former version of Google Play (in 2011). As soon as experiments and analysis is successfully conducted, signicant levels of pair-wise correlations are identied between some features for both datasets, which are further compared to achieve a generalized conclusion. Finally, cluster analysis is done to provide a similarity based recommendation system through probabilistic topic modeling method that can resolve Google Play's deciency upon app similarity.