Analisis Kinerja Metode Machine Learning (Orange Data Mining) Menggunakan Algoritma K-Nearest Neighbor, Random Forest dan Gradient Boosting dalam Prediksi Porositas Efektif Berdasarkan Data Well Log

Paris Paris (STT Migas Balikpapan)
Rohima Sera Afifah (STT Migas Balikpapan)
Baiq Maulinda Ulfah (STT Migas Balikpapan)
Bambang Sugeng (STT Migas Balikpapan)
Iin Darmiyati (STT Migas Balikpapan)

Jurnal Teknik Industri Terintegrasi (JUTIN)
Vol. 9 No. 3 (2026): July

Effective porosity (PHIE) is a critical petrophysical parameter in reservoir evaluation, yet its conventional determination through core analysis is constrained by high costs and limited data coverage, particularly in heterogeneous reservoirs with nonlinear log relationships. This study analyzes the performance of three machine learning algorithms, namely K-Nearest Neighbor (K-NN), Random Forest (RF), and Gradient Boosting (GB), in predicting PHIE from well logging data (Gamma Ray, Bulk Density, Neutron Porosity, Sonic Transit Time) using Orange Data Mining, with a 75% training and 25% testing split validated through Stratified 10-Fold Cross Validation. Results show Gradient Boosting achieved the best performance (R² = 0.828; RMSE = 0.012; MAE = 0.008), followed by Random Forest (R² = 0.812) and K-NN (R² = 0.754). The sequential boosting mechanism proved more adaptive in capturing nonlinear relationships between log parameters and effective porosity, offering an efficient tool for formation evaluation with reduced reliance on core data.