Tide prediction using soft computing techniques: Comparative analysis of ANN and ANFIS models for Cochin, India

Tidal prediction plays a vital ABSTRACT role in coastal engineering, navigation safety, environmental management, and disaster mitigation. Traditional harmonic analysis approaches require long-term datasets and often overlook meteorological influences, limiting their adaptability to dynamic conditions. This study evaluates two soft computing approaches, Artificial Neural Networks (ANN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS), for daily maximum water level prediction in Cochin, Kerala, using a decade-long dataset (2011–2021) of wind speed, wind direction, and surface pressure. ANN was implemented using the Non-linear Autoregressive with Exogenous Inputs (NARX) network architecture, while ANFIS was optimized with triangular membership functions. Both models were developed in MATLAB R2023a and evaluated using the Coefficient of Determination (R²) and Root Mean Square Error (RMSE). The ANN model achieved R² = 0.977 and RMSE = 0.013 for testing data, outperforming ANFIS (R² = 0.986, RMSE = 0.105). The maximum water level analysis further confirmed ANN’s superior accuracy, making it a robust alternative for real-time coastal flood risk management.