Classifying Legendary Pokémon with SF-Random Forest Algorithm

Journal: Journal of Information Systems and Informatics

Publication: 2024-09-23 · Vol. 6 No. 3 · pp. 1852–1871

DOI: 10.51519/journalisi.v6i3.859

Abstract

Here’s an improved version of the abstract with better articulation: Accurate classification of legendary Pokémon is essential due to their distinct characteristics compared to regular Pokémon, impacting various domains such as research, gaming, and strategy development. This study employs the SF-Random Forest algorithm, an advanced variant of Random Forest, designed to effectively handle data heterogeneity and complexity. The dataset comprises 800 Pokémon samples, including attributes like type, base stats (HP, Attack, Defense, etc.), and other relevant features. To address the inherent imbalance between legendary and non-legendary Pokémon, the data preprocessing phase includes outlier removal, handling of missing values, normalization through Min-Max Scaling, and class balancing using the SMOTE (Synthetic Minority Over-sampling Technique) method. The preprocessed data is then used to train the SF-Random Forest model, with performance evaluated using metrics such as accuracy, precision, recall, and F1-score. The results reveal that SF-Random Forest achieves perfect scores across all metrics, demonstrating 100% accuracy, precision, recall, and F1-score. This highlights the algorithm's superior ability to identify key features and manage data imbalance compared to traditional classification methods. The study underscores the efficiency and robustness of SF-Random Forest as a classification tool, paving the way for the development of more advanced classification systems applicable to various fields requiring complex pattern recognition.