Predicting water allocation trade prices using a hybrid Artificial Neural Network-Bayesian modelling approach

Abstract

This paper proposes an integrated (hybrid) Artificial Neural Network-Bayesian (ANN-B) modelling approach to improve the accuracy of predicting seasonal water allocation prices in Australia's Murry Irrigation Area, which is part of one of the world's largest interconnected water markets. Three models (basic, intermediate and full), accommodating different levels of data availability, were considered. Data were analyzed using both ANN and hybrid ANN-B approaches. Using the ANN-B modelling approach, which can simulate complex and non-linear processes, water allocation prices were predicted with a high degree of accuracy (RBASIC = 0.93, RINTER. = 0.96 and RFULL = 0.99); this was a higher level of accuracy than realized using ANN. This approach can potentially be integrated with online data systems to predict water allocation prices, enable better water allocation trade decisions, and improve the productivity and profitability of irrigated agriculture.