Comparison of Naive Bayes Algorithms and Decision Tree for Classifying Hero Fighter Items in the Mobile Legends

Abstract

The classification of hero fighter items in the Mobile Legends Game is a significant challenge due to the complexity of features and the variety of strategies employed by players. This study aims to develop an effective classification model using the Naïve Bayes and Decision Tree algorithms and compare the performance of these two algorithms. The dataset used in this study was obtained from item recommendations during live Gameplay, community sources such as forums, and Game guides. This dataset contains relevant information to support the classification of items for hero fighters, such as hero attributes, roles, and enemy types. The model training process was conducted using the scikit-learn library, with data split into 80% for training and 20% for testing. The study results show that the Decision Tree algorithm consistently delivers better performance than Naïve Bayes. In the evaluation using accuracy metrics, Decision Tree achieved an accuracy rate of 84.78%, significantly higher than Naïve Bayes, which only reached 45.65%. Furthermore, the precision, recall, and f1-score metrics for Decision Tree also showed superior results for almost all classes compared to Naïve Bayes. Based on these findings, the Decision Tree algorithm is recommended as a more suitable choice for classifying hero fighter items in the Mobile Legends Game.