Application of firefly algorithm-based support vector machines for prediction of field capacity and permanent wilting point

Soil field capacity (FC) and permanent wilting point (PWP) are significant parameters in numerous biophysical models and agricultural activities. Although these parameters can be measured directly, their measurements are quite expensive. The purpose of this study was to develop a hybrid Support Vector Machine (SVM) combined with Firefly Algorithm (FFA) techniques (SVM-FFA) to predict the FC and PWP using some easily available soil properties. The data consist of 215 soil samples collected from different horizons of soil profiles located in the East Azerbaijan provinces, North-west of Iran. Several important parameters, including the sand,silt, clay, bulk density, and organic matter content were used as inputs, while the soil FC and PWP were the output parameters. The predictions from the SVM-FFA model were compared with SVM and artificial neural network (ANN) models. The model results were compared with regard to root mean square error (RMSE), correlation coefficient (CC) and relative root mean square error (RRMSE). A comparison of models indicated that the SVM-FFA model predicted better than SVM and ANN models with RMSE = 2.402%, CC = 0.972, RRMSE = 7.677% for FC and RMSE = 1.720%, CC = 0.969, RRMSE = 5.512% for PWP in the training data set while RMSE = 2.873%, CC = 0.962, RRMSE = 8.745% for FC and RMSE = 1.935%, CC = 0.965, RRMSE = 10.619% for PWP were obtained in the testing data set.