<b>Comparative Evaluation of Ensemble ANN, GPR, and SVM Models for Predicting the Gross Calorific Value of Coal from Nigeria for Industrial Fuel</b> <b>Selection</b>

Accurate determination of the gross calorific value (GCV) of coal is essential for fuel quality assessment, blending, pricing, and industrial energy management. Conventional laboratory determination using bomb calorimetry, although highly accurate, is time-consuming, costly, and unsuitable for rapid large-scale evaluation. This study investigates the applicability of ensemble machine learning techniques for predicting the GCV of coal from Nigeria using readily obtainable proximate analysis parameters, namely air-dry moisture (ADM), residual moisture (RM), volatile matter (VM), and ash content (A). Three ensemble models, comprising ensemble artificial neural network (ensemble-ANN), ensemble Gaussian process regression (ensemble-GPR), and ensemble support vector machine (ensemble-SVM), were developed by aggregating three independently trained base learners to enhance prediction stability and reduce model variance. A total of 350 laboratory-tested coal samples were randomly partitioned using a hold-out split method into a calibration dataset comprising 300 data points to develop the models, and a verification dataset comprising 50 data points to compare the model performance.