ANALISIS KOMPARATIF KINERJA ALGORITMARANDOM FOREST DAN NAÏVE BAYES DALAM KLASIFIKASI SKALA INDUSTRI DI KABUPATEN KUBU RAYA

Abstract

The rapid growth of the industrial sector in Kubu Raya Regency has generated large volumes of company data, making manual classification increasingly difficult. This study compares the performance of the Random Forest and Naive Bayes algorithms in classifying industrial scale based on location, business type, and production capacity. The dataset consists of 591 industrial company records with an imbalanced class distribution. The research process includes data preprocessing, model training, testing, and performance evaluation using the Confusion Matrix, Accuracy, Precision, Recall, F1-Score, and AUC-ROC metrics implemented in Python through Google Colab. The results show that Random Forest outperformed Naive Bayes, achieving 95.80% accuracy, a macro-average F1-score of 60.51%, a macro-average AUC-ROC of 94.50%, and a five-fold cross-validation accuracy of 95.34%. In contrast, Naive Bayes achieved only 45.38% accuracy and a macro-average AUC-ROC of 71.33%. Statistical testing confirmed a significant performance difference, indicating that Random Forest is more effective for industrial-scale classification despite challenges in identifying the minority middle-scale class.