Comparisons of Tidal Prediction Analysis by Using Adaptive Neuro Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN)

Riau's coastal area such as Tanjung Buton, Region of Siak, is one of the maritime territory in Indonesia which has a quite huge tidal and the difference between flooding and ebbing reaches 3 - 4 meters. There are couple methods which had been used to forecast the tidal of the marine zone in order to determine the better solution to anticipate the occurred tidal problems. Soft computing method is the newest prediction model which is the most common-used method and based on knowledge, master system, fuzzy logic, artificial neural network (ANN), and genetics algorithm. The used data to create ANFIS model and ANN is the tidal data of sea water on Tanjung Buton in 2004 and supported by addition data from tidal data on Tanjung Motong in 2013 and Buatan in 2009. The simulations was made by three combinations, which are 1st simulation (70 : 30), 2nd simulation (65 : 35), and 3rd simulation (60 : 40). Those ratios were the number of data's percentage for calibration (learning) process, and verification (testing) process. Henceforth, ANFIS will forecast the tidal of those each data source and analyzes the ANFIS model's error in predicting tidals. The result of this research has indicated that the 1st simulation (70 : 30) was the best simulation compared to the 2nd and 3rd simulation because it produced the lowest error value. ANFIS method produced correlation value of 0.87208 whereas the ANN method produced the R score of 0.8766. Overall, the ANFIS model produced a better tidal prediction than ANN method did. The comparison of average relative error on ANFIS model for Tanjung Buton's data was only 11.8% whereas the average error of ANN model reached 12.09%.