Android Application Market Prediction Based on User Ratings Using KNN

Abstract

Google play store is a digital distribution service operated and developed by Google. It provides various applications that can be downloaded directly on Android devices, allowing users to browse and download their desired applications. When they search for an app, users can see the list of applications with the name and its rating on its side. It is helping users to choose the best application based on the application rating. Developers use their strategy in naming their applications attractively to increase their rating. In this paper, we create a market prediction modeling based on user rating for Android apps on Google Play Store throughout 2019 using the KKN algorithm. After that, we perform a feature engineering and an analysis using exploratory data analysis to find information in the dataset. By doing k=1-20 iteration, we can determine the best k value in KNN. We used 5-fold cross-validation to evaluate model, which is created with 88.68% accurate result, a recall of 87.46%, a precision of 89.5%, and an RMSE of 0.213. The generated model has a better performance compared to another algorithm, with a 10% increased accuracy.