Fuzzy expert system for ectomycorrhizal fungi growth response effectiveness prediction on forestry tree

Abstract

The growth response effectiveness of ectomycorrhizal fungi is influenced by environmental factors and compatibility with the host plan. An ectomycorrhizal fungi has its own growing characteristics including its level of effectiveness and interaction with the different physiology of the host plant. Thus, experience and expertise are required to make an accurate prediction of the effectiveness of ectomycorrhizal fungi. The objective at this study is designing an expert system model that can predict the effectiveness of ectomycorrhizal fungi quickly, accurately, and easily. The model is developed using a combination of decision trees and adaptive neuro fuzzy inference system (ANFIS). The process is divided into two stages. In the first stage, decision tree with classification and regression tree (CART) algorithm is used as feature selection procedure to select the most important variables. In the last stage, ANFIS is used to predict the growth response effectiveness of ectomycorrhizal fungi. The following process are showing how CART algorithm works, first the process of growing the tree based on the training data, second the process of pruning the tree which is based on the minimum cost-complexity principle. The CART results to identify the structure and parameter of ANFIS . The ANFIS model uses hybrid algorithm to do training and testing. The model performance is assessed by looking the correlation between the testing data with model output and mean absolutely percentage error (MAPE). The correlation model results are 0.998, 0.997, and 0.999 for Ekto1, Ekto2, and Ekto3 respectively. MAPE for Ekto 1 is 1.47%, Ekto 2 is 1.29% and 3.40% for Ekto3. Using the same model, performance tests on mycorrhizal experts are conducted as well. Keywords: expert system, fuzzy, decision trees, CART, ANFIS, effectiveness, ectomycorrhizal.