Statistics and neural networks for approaching nonlinear relations between wheat plantation and production in Queensland of Australia

An accurate prediction of wheat production in advance would give wheat growers, traders, and governmental agencies a great advantage in planning the distribution of wheat for business and consuming purposes. Traditional approach in dealing with such prediction is based on time series analysis through statistical or other intelligent means. These time-series centric methods treat the historical data as sequences of continuous events, and assume that the most recent sequence is more important than the earlier ones in forecasting. However, such analysis concerns little about the factors that cause the appearances of the events. In wheat production prediction, factors, such as the total plantation area, variations in rainfall and temperature, and levels of fertilization and disease occurrence, all make contributions to the harvest. In this paper, treating the historical wheat data in Queensland over 130 years as non-temporal collection of mappings between wheat plantation area and production, we use correlation analysis and neural network techniques to reveal whether significant nonlinear relations exist between these two factors. If such nonlinear relations exist, comparisons are then made to identify the best possible solution that can be used for predicting wheat production with respect to the plantation area. Our investigation indicates that similar study has not been published yet. Our analysis demonstrates that a power correlation, a third-order polynomial correlation, and a three layer multilayer perceptron model are all of significance, but it is the multilayer perceptron model that is capable of producing accurate prediction.