Predicting Mobile App Success Using a Robust Hard Voting Ensemble Learning Approach

Abstract

Apps have become an inseparable part of our daily lives. There are different kinds of mobile apps on the market. Google Play Store apps development is one of the most enticing and consumer-friendly development paradigms for mobile apps. On the other hand, the paradigm is still in its early phases and does not address critical issues such as an app’s success and failure. A considerable number of mobile apps do not acquire a good solution, squandering stakeholders’ time and effort. Therefore, predicting the success of a new app will be helpful for developers. This research proposes an ensemble learning-based approach for predicting the success of the mobile app. For this purpose, the app’s important attributes (rating, number of installs) can be selected from the dataset. Dataset can be preprocessed using NLP (Natural Language Processing) technologies and perform Data Analysis. These selected features were then deployed to several ML (Machine Learning) algorithms — Deci- sion Tree Classifier, Random Forest Classifier, K - Nearest Neighbour Classification, Gradient Boosting Classifier, and Light Gradient Boosting Classifier. Finally, an ensemble model pro- poses to predict a new app’s success. Our suggested model outperforms, with an accuracy of 96.772239%.