Predicting Aluminum Oxide Grade

Abstract

AbstractThis chapter explains the application of Bayesian model averaging and individual models for estimating aluminum oxide grade. The latitude, longitude, and depth of boreholes are used for estimating aluminum oxide grade. In the first level, the multi-layer perceptron (MLP)–particle swarm optimization (PSO), MLP–salp swarm algorithm (SSA), MLP–naked mole rate algorithm (NMR), MLP–genetic algorithm (MLP-GA), and MLP are used for estimating aluminum oxide grade. In the next level, the outputs of these models are inserted into Bayesian model averaging (BMA). The mean absolute errors (MAE) of the BMA, MLP-NMR, MLP-SSA, MLP-PSO, MLP-GA, and MLP models were 10.12, 10.98, 11.12, 12.23, 14.45, and 15.56 at the testing level. The Nash Sutcliffe efficiencies (NSE) of the BMA, MLP-NMR, MLP-SSA, MLP-PSO, MLP-GA, and MLP models were 0.94, 0.92, 0.89, 0.86, 0.84, and 0.82 at the testing level.KeywordsMultilayer perceptron (MLP)Particle swarm optimization (PSO)Salp swarm algorithm (SSA)Estimating ore grade