Microplastics in Drinking Water

Dataset Overview

Dataset Name

• The dataset is named “Microplastics in Drinking Water,” with the specific file called “Microplastics Sample Data (wide).”

Dataset Source

• Released by the California State Water Resources Control Board, accessible via: Microplastics in Drinking Water.

Dataset Content

• Each row represents a water‑sample record with associated information.
• Key columns include microplastic material and type (content per sample), color, tap vs. bottled water, sampling location, and approximate coordinates.
• The project focuses on PE (polyethylene); other “material” columns are removed.

Data Processing

• The original dataset had over 100 columns; columns with fewer than 40 values were removed.
• Additional cleaning removed unnecessary columns such as Sample_ID and handled all NAN or Present values.
• Samples from Chinese reservoirs with extreme values were excluded.

Dataset Usage

• Random Forest, k‑NN regression, and Decision Tree regression models were employed for prediction.
• Model evaluation indicated Decision Tree regression performed best, though its predictive power is limited by sample size and data quality.

Dataset Limitations

• The dataset suffers from many missing values and mismatched data types; after cleaning, only about 60 samples remain usable.
• Updated continuously since 21 July 2022, but current data reliability and standardization are insufficient for robust predictive modeling.

Conclusion

• Despite testing multiple models, the dataset’s quality prevents reliable predictions of drinking‑water safety based on microplastic content. Further data collection and standardization are required.