A method for estimating productivity of Zoysia japonica using neural network

Abstract

Two production models, one using a neural network and the other using a multiple regression analysis, were constructed based on data of monthly dry matter production under various climatic conditions for Zoysia japonica. These models were then evaluated for practical use by comparing conformity of estimated data with actual harvest data. Using 105 items of harvest data with Zoysia japonica for three years, these models were compared for the estimation accuracy each other. The results clearly showed that the neural network had a higher coefficient of correlation between estimated data and harvest one. Accordingly, the neural network model showed the better estimation for the productivity.