Water quality index modeling using random forest and improved SMO algorithm for support vector machine in Saf-Saf...

Area of Science:

• Environmental Science
• Artificial Intelligence
• Water Resource Management

Background:

• Surface water quality assessment is crucial for effective river water resource management.
• The Water Quality Index (WQI) serves as a key indicator for classifying surface water quality.
• Predictive modeling aids in understanding and managing water quality dynamics.

Purpose of the Study:

• To develop and evaluate hybrid artificial intelligence models for predicting WQI.
• To compare the predictive performance of Sequential Minimal Optimization-Support Vector Machine (SMO-SVM) and Random Forest (RF) models.
• To identify optimal input parameter combinations for accurate WQI prediction in the Wadi Saf-Saf river basin.

Main Methods:

• Construction of a hybrid SMO-SVM model and a Random Forest (RF) model for WQI prediction.
• Utilized fifteen water quality parameters as input variables, including BOD, pH, DO, EC, TDS, and temperature.
• Evaluated model performance using statistical metrics and sensitivity analysis across different input data combinations.

Main Results:

• The RF model demonstrated higher accuracy in WQI prediction than the SMO-SVM model for both training and testing datasets.
• Optimal predictive performance was achieved using thirteen input parameters (R² testing = 0.82, RMSE testing = 5.17).
• A subset of five parameters (pH, EC, TDS, T, saturation) also yielded high precision (R² test = 0.81, RMSE testing = 5.55).

Conclusions:

• The Random Forest model shows significant improvement over existing tools for WQI prediction.
• RF model's ability to maintain high accuracy with reduced input variables enhances its practical applicability.
• Findings suggest RF is a promising tool for efficient and accurate water quality monitoring and management.