Analysis of accuracy parameters of ANN backpropagation algorithm through training and testing of hydro-climatology data based on GUI MATLAB
Syaharuddin Syaharuddin, Dewi Pramita, Toto Nusantara, Subanji Subanji, Habib Ratu Perwira Negara · IOP Conference Series Earth and Environmental Science · 2020
Abstract The authors have developed a GUI Matlab to simplify the process of predicting Hydro-climatology data using ANN Back Propagation method. Five data for training, testing, and prediction were used. The data, i.e. rainfall, air humidity, duration of shine, temperature, and wind speed are taken from the last ten years with matrix input size m x n. Each data is trained 21 times using a combination of the activation functions (logsig, tansig, and purelin) and training methods (traingda, traingdx, and trainrp). The result of the training data was that the logsig function and trainrp on each layer are the best formulas in conducting training, testing, and predictions with an accuracy of 99.71%. This result is obtained from parameter settings including epochs of 1000, learning rate of 0.7, goal error of 0.0001, and training steps of 1.