Evaluation of the Hybrid Artificial Neural Network-Coati Optimization Algorithm (ANN-COA) Model for Predicting Saffron Water Demand Using Limited Climatic Parameters

Accurate estimation of saffron water demand is essential for sustainable water resource management in saffron-growing regions. This study examines the optimization of the Artificial Neural Network (ANN) model for predicting saffron water demand using the hybrid Coati Optimization Algorithm (COA). The performance of the ANN-COA model was compared with ANN, ANN-GA, ANN-PSO, ANN-MFO, Quadratic Regression (QR), Tree Regression (TR), and Pattern Regression models. Input data included temperature (minimum, maximum, average), wind speed, relative humidity, net radiation, and day of the year. The results showed that under conditions using all climatic parameters, the ANN-COA model achieved an R² of 0.995 and a Mean Squared Error (MSE) of 0.0001 for the Mashhad station, and an R² of 0.973 and MSE of 0.0005 for the Birjand station, indicating acceptable accuracy in predicting saffron water demand. Additionally, under conditions with limited climatic parameters, the ANN-COA model, using maximum temperature, wind speed, and day of the year, exhibited the best performance in predicting saffron water demand. Based on the findings of this research, hybrid neural network models show superior accuracy in estimating saffron water demand with minimal climatic parameters compared to other data mining models.