Unsupervised Learning Detects Heavy Metal Anomalies at Ghana Waste Sites
A study published on arXiv applies unsupervised machine learning to detect anomalous heavy metal contamination at unregulated waste disposal sites in Ghana's Central Region. Researchers analyzed concentrations of arsenic, cadmium, chromium, copper, mercury, nickel, lead, and zinc in 78 soil samples from twelve waste sites and residential controls. Isolation Forest and PCA reconstruction error each flagged 12 anomalous samples (15.4%), while DBSCAN found none. A consensus approach identified six robust anomalies (7.7%), all concentrated at a single site (S3). These anomalies showed approximately 70–80% higher mean Hazard Index values, indicating elevated health risks. The framework demonstrates potential for environmental risk assessment in data-scarce regions.
Key facts
- Study applies unsupervised ML to detect heavy metal anomalies in Ghanaian soils
- 78 samples from twelve waste sites and residential controls in Central Region, Ghana
- Eight metals analyzed: As, Cd, Cr, Cu, Hg, Ni, Pb, Zn
- Isolation Forest and PCA each identified 12 anomalous samples (15.4%)
- DBSCAN detected no noise points
- Consensus approach isolated six robust anomalies (7.7%) at site S3
- Anomalies had 70–80% higher mean Hazard Index values
- Published on arXiv with ID 2604.27102
Entities
Institutions
- arXiv
Locations
- Ghana
- Central Region, Ghana