A Hybridized Forecasting Model for Metal Commodity Prices: An Empirical Model Evaluation

Abstract

Appropriate decision on perfect commodity prediction under market’s constant fluctuations intensifies the need for efficient methods. The main objective of this study is to apply different optimization algorithms such as conventional particle swarm optimization (PSO), bat algorithm (BAT) and ant colony optimization (ACO) algorithms on back propagation neural network (BPNN) to enhance the accuracy of prediction and minimize the error. In this paper, a model has been proposed for volatility forecasting using PSO algorithm to train the BPNN and predict the commodities’ closing price. The proposed PSO-BPNN model is considered as best forecasting model compared to BPNN, BAT-BPNN and ACO-BPNN. The experiment has been carried out upon five publicly available metal datasets (gold, silver, lead, aluminium, and copper) to forecast the price return volatility of those five metals challenging the effectiveness. Here, three technical indicators and four filters, such as; moving average convergence/divergence (MACD), williams %R (W%), bollinger (B), least mean squares (LMS), finite impulse response (FIR), Kalmanand recursive least square (RLS) have been applied for providing an additional degree of freedom to train and test the classifiers. From the experimental result analysis it has been found that the proposed PSO-BPNN produces promising output while comparing with BPNN, BAT-BPNN and ACO-BPNN.