Hybrid Intelligence for Coastal Pattern Recognition: ANFIS-Based Prediction of Multi-Level Beach Cusp Spacing

Valipour, Azadeh; Shirgahi, Hossein

doi:10.22034/ijcoe.2025.523922.1171

Hybrid Intelligence for Coastal Pattern Recognition: ANFIS-Based Prediction of Multi-Level Beach Cusp Spacing

Articles in Press

Document Type : Original Research Article

Authors

Azadeh Valipour ¹

Hossein Shirgahi ²

¹ Department of Marine Science and Technology, Jouybar Branch, Islamic Azad University, Jouybar, Iran

² Department of Computer Engineering, Jouybar Branch, Islamic Azad University, Jouybar, Iran

10.22034/ijcoe.2025.523922.1171

Abstract

Given the necessity of understanding coastal dynamics and predicting erosion in coastal areas, this research employs the Adaptive Neuro-Fuzzy Inference System as a machine learning model for predicting beach cusp spacing and subsequently optimizing coastal management. ANFIS has the advantage over other techniques to model the more complex nonlinear relationships involved in beach cusp formation. It learns from existing data, adapts to new information varying with coastal hydrodynamic conditions, and gives back to the user interpretability by showing underlying rules dictating cusp dynamics. This paper discusses implementing ANFIS to predict beach cusp spacing and examines its performance against a neural network. In this way, different ANFIS configurations were tested, and the effect of membership functions on the performance of the fuzzy system was investigated. The main outcomes of this research indicate that even though optimized Artificial Neural Network (ANN) models perform reasonably well for the upper beach face, the optimized ANFIS (trimf) model performs better in accuracy and stability for the middle and lower beach faces. This study effectively highlights the importance of selecting the optimal model tailored to each beach section's specific conditions and characteristics.

Keywords

Adaptive Neuro-Fuzzy Inference System

Machine learning

Coastal management

Membership functions

Neural network

Subjects