Prediction of weighted entropy water quality index for agricultural uses using simple and hybrid artificial intelligence methods

Introduction
Water quality is a crucial environmental issue affecting health and ecosystems. Assessing irrigation water quality is vital for making informed decisions about water consumption and reducing crop pollution. Water quality forecasting plays a crucial role in evaluating and identifying possible changes in water quality. Artificial intelligence (AI) is a powerful tool that can predict and monitor water quality. AI models can recognize complex patterns in environmental data and provide accurate predictions. The sub-indices of the agricultural water quality index were calculated, and the weight of the sub-indices was determined using Shannon's entropy theory. Simple and combined AI algorithms were used to predict the weighted entropy water quality index.

Methodology
This study focuses on the Karun River, located in the Khuzestan Province of Iran, specifically between Gatond, Ahvaz, and Sulaymaniyah stations. The province is divided into two regions, namely, mountainous and plain, with the plain region accounting for 60% of the province. In this study, the irrigation water quality index (IWQI) was used to determine the suitability of water for irrigation. The IWQI was calculated using five parameters, including EC, Na⁺, Cl⁻, SAR, and HCO^-₃. The weight of these parameters was determined using the Shannon entropy technique. Based on the IWQI values, water for irrigation was categorized into five classes. The study used data from the period of 1370 to 1400. This study developed a predictive model using the decision tree algorithm to determine agricultural consumption's weighted entropy water quality index. To address overfitting issues, artificial intelligence techniques were incorporated. Pearson correlation analysis was used to select the important input parameters. The combined decision tree and neural network algorithm were used to test the model's efficiency. The selection of important input parameters can enhance the accuracy and efficiency of the model, which is a crucial aspect of better artificial intelligence algorithms.

Results and discussion
Electrical conductivity measures the ionic strength of irrigation water and determines its suitability for agriculture. Studies have shown that high electrical conductivity and sodium absorption ratio can negatively impact soil and plant health. In the studied area, electrical conductivity values range from 428 to 4846 ds/m, with 9.36% of the samples exceeding 3000. SAR values range from 0.96 to 11.64, with 21.91% of the samples in the illegal category. Figures 2 and 3 show the distribution of irrigation water quality index values in five categories. Low water quality is due to the high electrical conductivity and SAR, which can be attributed to various factors, including human activities such as dam construction. The study area's irrigation water quality index values are classified into five categories, with most samples having moderate restrictions. The water quality in Gatund station is better than Sulaymaniyah station, which has the lowest quality due to various factors The prediction results of the water quality index showed that the combined algorithm of decision tree and neural network has a higher ability than the simple algorithm of decision tree and neural network.
The prediction results of the water quality index showed that the combined algorithm of decision tree and neural network has a higher ability than the simple algorithm of decision tree and neural network.
Hybrid models can effectively depict patterns and relationships in water quality data, resulting in more reliable predictions. Combining different algorithms and techniques in hybrid models increases their predictive capabilities and makes them more suitable for water quality prediction tasks. One of the effective ways to reduce the cost and time of determining the water quality index is to reduce the number of unimportant input parameters. The results of this research showed that bicarbonate ion can be removed due to its low effectiveness in the present study. The results of the best algorithm before and after reducing the input parameters are shown in the beloew figure and table.

Conclusions
AI techniques are used to predict environmental processes. Multi-criteria decision-making improves water quality assessment accuracy. Simple and combined AI algorithms efficiently estimate the weighted entropy water quality index for agriculture. Combined algorithms are more effective.