This project focuses on analyzing and predicting water quality levels using various parameters that define its potability. Through an in-depth dataset and employing machine learning models, we aim to achieve accurate predictions to ensure water safety and quality.
The dataset comprises several parameters vital for assessing water's quality and safety for consumption:
- pH Value: Indicator of water's acid-base balance, within WHO recommended range of 6.5 to 8.5.
- Hardness: Measurement of calcium and magnesium salts, which determine water's ability to precipitate soap.
- Solids (TDS): Reflects water's mineralization with a desirable limit set at 500 mg/L for drinking water.
- Chloramines: Disinfectants used in water, safe up to 4 mg/L.
- Sulfate: Naturally occurring substances with concentrations in freshwater typically ranging from 3 to 30 mg/L.
- Conductivity: Indicator of water's ionic concentration with a standard limit not exceeding 400 μS/cm as per WHO.
- Organic Carbon: Measures carbon in organic compounds in water, with EPA standards set for treated and source water.
- Trihalomethanes (THMs): By-products of chlorine treatment, safe up to 80 ppm.
- Turbidity: Measure of water's clarity, with WHO standards recommending values below 5.00 NTU.
- Potability: Binary indicator of whether water is safe (1) or not (0) for consumption.
The project entails a comprehensive approach to predict future water quality levels accurately, covering:
- Data collection
- Preprocessing
- Feature engineering
- Modeling phases
We utilized Random Forest, XGBoost, and LightGBM models in a streamlined pipeline, achieving:
- An 83% accuracy rate in forecasting water quality parameters.
Our project demonstrates the effectiveness of machine learning techniques in predicting water quality, contributing towards ensuring the safety and potability of water.
Contributions are welcome to improve the accuracy of predictions and expand the dataset. Please refer to the contributing guidelines for more details.
This project is licensed under the MIT License - see the LICENSE file for details.
For any inquiries or contributions, please contact us at sudesokin@gmail.com.