Water Quality Prediction Using Random Forest Classifier: An Analysis of Chemical Attributes and Their Feature Importance

This work focuses on the use of machine learning algorithms to estimate water quality based on several environmental variables including $\boldsymbol{p H}$, Iron, Nitrate, and other chemistries. With the help of a large data set, we had tested the proposed parameters and their impact towards water quality with the help of Random Forest Classifier. In the model, it was possible to reach the level of 83.17 accuracy. Exploratory feature importance analysis for the selected model provided insights that showed that Manganese, Turbidity, and Odor were dominating indicators for water quality. A test of discriminant validity through a Receiver Operating Characteristic (ROC) curve and an Area Under Curve (AUC) of 0.76 confirmed the model’s good discriminant validity in the differentiation of the water quality classes. Further, the clustering approaches were employed to analyze data clusters using pH, Iron, and Nitrate parameters to understand the inherent grouping of the data set. Some examples were $2 D$ and 3D scatter plots, including the use of these clusters to show the extent of the impacts in identifying the underlying trends in a given set of data and thus provide an enhanced capability of coming up with hypotheses of the phenomenon under investigation. The conclusions derived from this study present the essential factor influencing water quality and the effectiveness of machine learning models for environmental applications. This work paves the way for further exploration of predictive analytics in sustainable water management practices.