Evolutionary Learning of Fuzzy Rules and Application to Forecasting Environmental Impact on Plant Growth

Abstract

: Prediction of plant growth and yield is one of the essential tasks that enables growers of food and agricultural products to effectively manage their crops. In this paper, a hybrid evolutionary/fuzzy machine learning approach is introduced where a genetic algorithm is deployed to learn the optimum membership functions of relevant fuzzy sets and a knowledge base of fuzzy rules. This hybrid approach is then used to build a model which determines how ozone and carbon dioxide levels in the atmosphere affect plant growth by predicting the basal width growth of a plant. The hybrid forecasting model was tested on a data set collected from soybean fields and proved to be an extremely accurate and robust fuzzy predictor. It was able to predict the basal width growth of the plant with an average of 0.19% relative absolute value error.