Exploratory data analysis and success prediction of Google Play Store apps

Abstract

Mobile app distribution platform such as Google play store gets flooded with several thousands of new apps everyday with many more thousands of developers working independently or in a team to make them successful. With immense competition from all over the globe, it is imperative for a developer to know whether he is proceeding in the right direction. Unlike making a movie where presence of popular celebrities raise the probability of success even before the movie is released, it is not the case with developing apps. Since most Play Store apps are free, the revenue model is quite unknown and unavailable as to how the in-app purchases, in-app adverts and subscriptions contribute to the success of an app. Thus, an app’s success is usually determined by the number of installs and the user ratings that it has received over its lifetime rather than the revenue it generated. In this thesis, on a smaller scale, we have tried to perform exploratory data analysis to dive deeper into the Google Play Store data that we collected, discovering relationships with specific features such as how the number of words in an app name for instance, affect installs, in order to use them to find out which apps are more likely to succeed. Using these extracted features and the recent sentiment of users we have predicted the "success" of an app soon after it is launched into the Google Play Store.