A hybrid data-driven framework for diagnosing contributing factors for soil heavy metal contaminations using machine...

Area of Science:

• Environmental Science
• Data Mining
• Geochemistry

Background:

• Soil heavy metal (HM) contamination poses significant environmental risks.
• Accurate source apportionment is crucial for effective remediation but often hindered by data limitations.

Purpose of the Study:

• To develop a novel hybrid data-driven framework for diagnosing factors influencing soil HM contamination.
• To overcome limitations in identifying natural and anthropogenic HM sources in industrial regions.

Main Methods:

• A hybrid framework combining Naive Bayes (NB), Random Forest (RF), and Bivariate Local Moran's I (BLMI).
• Utilized multi-source big data, including Baidu Point of Interest data for enterprise classification.
• Optimized NB for identifying contaminating enterprises and RF for quantifying contributions.

Main Results:

• Successfully classified medium industry types and identified 250 contaminating enterprises.
• Quantified the contributions of nine factors to As, Cd, and Hg concentrations using RF.
• Generated spatial clustering maps revealing interactions between HM concentrations and key factors.

Conclusions:

• The proposed framework effectively identifies and quantifies sources of soil HM contamination.
• Provides rich information on industry types, contribution rates, and spatial distributions of pollutants.
• Offers a powerful tool for environmental management and pollution control in industrialized areas.