An Approach To Predict Software Application Success Using Voting Ensemble Method

Abstract

Mobile app distribution platform consisting of Google play store and Apple Store gets covered with several hundreds of new apps every day with many more enthusiastic developers working independently or in a crew to make them successful. With huge competition from all over the world, it is vital for a developer to recognize if he is proceeding in the proper direction or not. It is not like making a film wherein presence of famous celebrities increase the chance of success even earlier than the movie is released, it is not the case with developing apps. Since maximum Play Store apps are free, the revenue version is pretty unknown and unavailable as to how the in-app purchases, in- app advertisements and subscriptions make a contribution to the fulfillment of an app. Thus, a software’s success is normally decided through the wide variety of installs and the star ratings that it has acquired over its lifetime in place of the sales it generated. So in order to test if the app is assembly the expectancy of human beings we need a software that will check that apps evolved are successful or not. The framework for predicting success of software application is proposed in this paper. It is a software which will provide the achievement of app/software program relying not only on number of install and star rating but will consider all the factors of an application description and customers reviews. The exploratory data analysis to jump in deeper into the Google Play Store information is performed. The relationships with specific functions inclusive of how the wide variety of phrases in an app call for instance, affect installs are used to use them to find out which apps are much more likely to succeed. Using those extracted functions and the of sentiment of customers the proposed method will predict the “success” of an application using Google Play Store Data. The algorithm applied are Support Vector Machine, K-Nearest Neighbour, Decision Tree and Random Forest. In order to improve accuracy the Voting Ensemble technique is applied in proposed method. The accuracy of various algorithm is compared to judge the performance of proposed method. Keywords : hyperplane, dataset, Support Vector Machine, Exploratory Data Analysis, voting ensemble