MACHINE LEARNING APPROACHES FOR WATER QUALITY INDEX PREDICTION IN PABA UPAZILA, RAJSHAHI: A COMPARATIVE STUDY OF MINITAB AND ADVANCED ALGORITHMS

Predicting the Water Quality Index (WQI), which provides communities and policymakers a measurable indicator of water quality, is crucial for efficient environmental management. The purpose of this study is to investigate numerical models for predicting the WQI values using Minitab (regression analysis) and machine learning algorithms namely Decision Tree (DTR), Random Forest (RGR), Stochastic Gradient Decent (SGD), and Support Vector Machine (SVR). This is accomplished by collecting surface and ground water from 200 locations in the Paba Upazila, Rajshahi and doing laboratory tests to determine the pH, turbidity, total dissolved solids and total solids to create an extensive dataset that reflects the water conditions in the area. The WQI is then computed using the parameters from the Brown et al. (1972) technique. According to the analysis, Minitab and SVR perform better than the others, obtaining strong classification metrics (93% accuracy, 0.94 F1-score) and remarkable prediction accuracy (r2 = 0.9503 for Minitab; r2 = 0.9443 for SVR). The intricate interactions between the several water quality indices in the study area are well captured by these models. With a data-driven strategy to monitoring and forecasting water quality in Paba Upazila, the findings offer significant insights for local water resource management. The results of the evaluation can provide a scientific basis for the conservation of the local aquatic environment, and the model created in this study can be used as a guide for similar water quality assessment work. This study also highlights the potential of integrating machine learning algorithms with statistical software such as Minitab for environmental monitoring applications, and it helps design customized solutions for water quality evaluation in comparable regions of Bangladesh.