Back To Projects

A Comprehensive Study of Machine Learning Models for the Prediction of pH in Fairfax County

Matthew N.

Water quality is drastically declining as bodies of water become more and more acidic due to global warming. Because of this, it is crucial to understand what factors determine pH, and how to predict pH values. In this study, we determine which features are most important in determining pH, and also what combination of features and machine learning models give us the best accuracy in the prediction of pH values. We demonstrate that pH values from the previous year and the percentage of rehabilitated sanitary sewer lines are the most important features in determining pH. We also demonstrate that Decision Tree Models and Random Forest Models perform the best across all features and feature subsets.


Water quality is drastically declining as bodies of water become more and more acidic due to global warming. Because of this, it is crucial to understand what factors determine pH, and how to predict pH values. In this study, we determine which features are most important in determining pH, and also what combination of features and machine learning models give us the best accuracy in the prediction of pH values. We demonstrate that pH values from the previous year and the percentage of rehabilitated sanitary sewer lines are the most important features in determining pH. We also demonstrate that Decision Tree Models and Random Forest Models perform the best across all features and feature subsets.

Explore More!

Matthew N.
Sarthak Kanodia

Related Projects

workspace_premium
Combating Climate Fake News Using NLP

As fake news becomes more prevalent across the US, important issues become harder to solve. One such issue is climate change, where climate misinformation has worsened viewer’s abilities to distinguish between fake information and real information. This project’s objective is to tackle climate misinformation using an artificial intelligence model.
Rayyan M.
Mentored by Philip Bell
Examination of Dragonflies (Pantala Flavescens): Characteristics, Identification and Migration Paths

This study explores the Pantala flavescens species' anatomy and mechanisms for long-distance migrations.
Alya K.
Mentored by West Foster
Predicting NH4 Levels for Corn Crop in Wisc

Ammonium (NH4), an organic matter that accumulates in the top portion of soil, can pose a serious risk to biodiversity. Using machine learning to construct regression models, NH4 levels can be predicted and therefore mitigated. In this paper we used linear, ridge, and lasso regressions. Through the evaluation of crop farming factors that contribute to the NH4 levels, it was concluded that NO3 and N2O have the most direct correlation to NH4. These factors yielded the best accuracy for regression models with the best performing model being a multiple feature linear regression which resulted in 60% accuracy. While certain measures did improve the model’s performance, outliers continuously worsened the results.
Julia S.
Mentored by Barbie Duckworth