IMPLEMENTATION OF DATA MINING ALGORITHM FOR PREDICTING POPULARITY OF PLAYSTORE GAMES IN THE PANDEMIC PERIOD OF COVID-19

Abstract

The existence of the COVID-19 virus makes everyone fill their time at home by doing various activities, one of them playing games on the phone. For the game to develop continuously, it needs an assessment that comes from the community and especially the game lovers themselves. This assessment is used to find out what category of game you want. Therefore the analysis is needed to determine the interests of game lovers by analyzing the popularity of a game. This research was conducted to predict the level of popularity of games in PlayStore applications to find out how many popular and unpopular games and the accuracy obtained with the C4.5 algorithm and Naive Bayes algorithm. The results obtained using the C4.5 algorithm showed 73 popular games and 12 unpopular games with an accuracy value of 85.83% with a precision of 85.83% and a recall of 100% and Naive Bayes showed 23 popular games and 62 unpopular games with an accuracy value of 80% with a precision of 96.11% and a recall of 81.01%. The evaluation results with the ROC curve show the AUC value using the Naive Bayes model of 0.776 and the C4.5 model of 0.500. Of the two models used, one of them is included in the classification of Good classification, namely the Naive Bayes algorithm model, because it has an AUC value between 0.80-0.90. While the C4.5 algorithm model is included in the Fair classification, has an AUC value between 0.70 - 0.80.