An improved adaptive neuro fuzzy inference system model using conjoined metaheuristic algorithms for electrical...

Area of Science:

• Environmental Science and Engineering
• Water Resource Management
• Computational Intelligence

Background:

• Accurate prediction of water quality parameters like electrical conductivity (EC) is vital for early pollution detection and effective water resource management.
• Electrical conductivity is a key indicator of water mineralization, influencing overall water quality assessment.

Purpose of the Study:

• To develop and evaluate an advanced hybrid model for precise monthly prediction of electrical conductivity (EC) in the Maroon River, Iran.
• To enhance the prediction accuracy of EC by integrating wavelet analysis with an adaptive hybrid optimization algorithm and a neuro-fuzzy inference system.

Main Methods:

• An Adaptive Hybrid of Differential Evolution and Particle Swarm Optimization (A-DEPSO) was employed for training the Adaptive Neuro-Fuzzy Inference System (ANFIS).
• Wavelet analysis was utilized to decompose the EC time series into sub-series, improving prediction certainty.
• The proposed Wavelet-ANFIS-A-DEPSO (W-ANFIS-A-DEPSO) model was compared against standalone ANFIS, LSSVM, MARS, GRNN, and their wavelet-integrated counterparts.

Main Results:

• The W-ANFIS-A-DEPSO model demonstrated superior performance in EC prediction, achieving a correlation coefficient (R) of 0.988, RMSE of 53.841, and PI of 0.485.
• This hybrid model significantly outperformed other tested models, including standalone ANFIS-DEPSO, improving RMSE by 80%.
• The Dmey mother wavelet integration within the W-ANFIS-A-DEPSO framework yielded the best predictive accuracy.

Conclusions:

• The W-ANFIS-A-DEPSO model offers a highly accurate and promising approach for simulating and predicting electrical conductivity in water bodies.
• This advanced modeling technique can serve as a valuable tool for water quality monitoring and management, aiding in pollution control and resource optimization.