Improvement accuracy of gradient boosting in app rating prediction on google playstore

Abstract

Google Playstore is a platform that provides various useful applications for smartphone users, especially Android users. But in reality, users are often faced with many choices of applications with various features and functions in the Play Store itself. Rating applications on the Google Play store help users evaluate and choose applications that suit their needs. The purpose of this research is to optimize accuracy in predicting app ratings on the Google Play store using the Gradient Boosting algorithm. This research uses publicly accessible data on the Kaggle platform. The research process includes data collection, pre-processing, Data Spliting, algorithm modeling, and model evaluation. Apart from using the Gradient Boosting algorithm, this research also applies and optimizes other algorithms such as XGBoost, KNN, Logistic Regression, Decision Tree, Random Forest, LightGBM, AdaBoost, and SVM to predict app ratings on Google Playstore. By implementing and optimizing these algorithms, this study succeeded in achieving an accuracy of 92.62%, with MAE 0.311, RMSE 0.467, and R-square 0.144 using the Gradient Boosting algorithm. This research contributes to the development of better prediction methods in the mobile application industry and provides new insights regarding the factors that influence app ratings on the Google Play store.