<![CDATA[This study aims to classify Pokémon types based on battle statistics using the Random Forest algorithm. The dataset used comes from the file pokemon_bw.csv, which contains information such as Pokédex number, name, type, abilities, and battle stat values (HP, Attack, Defense, Special Attack, Special Defense, Speed). Data preprocessing was carried out to clean and prepare the data, including primary type extraction, label encoding, feature selection, and feature standardization. After that, the dataset was split into training and testing data with an 80:20 ratio. The classification model was built using Random Forest with 100 decision trees and evaluated using accuracy, classification report, confusion matrix, and multiclass ROC Curve metrics. The results show that the model achieved an accuracy of 64.8%, with the best performance in the 'rock', 'steel', and 'dragon' classes, while the 'flying' and 'ghost' classes were still difficult to classify accurately. Confusion matrix. It shows that some types have quite significant misclassification errors, such as 'ground' which is often predicted as 'grass' and 'rock' which is often misclassified as 'steel'. ROC Curve evaluation also proves that most classes have an AUC above 0.80, indicating the model's ability to distinguish between classes. With this approach, this study provides an initial analysis of the potential for predicting Pokémon types based on battle statistics, which can be further developed through handling class imbalances or using other ensemble techniques.]]>
