Evaluating of performance artificial neural network with genetic algirithm and fuzzy logic in infiltration rate estimation of Ghezel Ozan river soils in Zanjan

Prediction of the rate of water infiltration into the soil and its influencing factors is the most important issues in the management of soil resources. In this study, the infiltration rate of the soil was estimated using 3 methods of artificial neural networks (ANN), Adaptive neuro fuzzy inference system (ANFIS) and genetic algorithm neural networks (GA-NN). The input data to all three methods in 8 information structurs including physical and chemical characteristic of soils from 60 sampling points in Ghezel Ozan river of Zanjan province. Evaluation results based on the statistics model of, root mean square error (RMSE), correlation coefficient (r) and mean absolute error (MAE) and the coefficient of performance. Statistics RMSE, MAE, and r in the best GA-NN architecture, respectively, 0.27, 0.15 and 0.99. This algorithm improved results 82.2% and 85.2 % compared to ANFIS formation and 48.1 % and 21.1% compared to ANN with 1-4-9 structure. Similarly, the coefficient of performance of the GA-NN structure, 0.996, ANFIS 0.863 and ANN 0.984, respectively, which reflects the performance of GA-NN. The Comparison of results of T- Student test from each of the three methods is in the 0.05 and 0.01, indicating no significant difference between the 8 structures. So, in 8 structure with three neurons including the percentage of soil particles.