A Predictive Framework for Marine Microplastic Pollution using Machine Learning and Spatial Analysis

Microplastic pollution in marine environments presents a critical ecological threat, affecting biodiversity and human health through bioaccumulation and ecosystem disruption. Traditional monitoring methods are labour-intensive, geographically constrained, and lack scalability for global assessments. This study proposes a predictive framework that integrates machine learning with geospatial analysis to forecast microplastic pollution density across oceanic regions. A dataset of global ocean samples was enhanced using domain-specific features such as Annual Trend Score (ATS), Regional Microplastic Concentration (RMC), and Plastic Pollution Index (PPI). Four models— Gradient Boosting, Random Forest, Linear Regression, and Support Vector Regression—were evaluated. Gradient Boosting achieved the best balance of accuracy and generalization, with an R2 of 97.17% and MAE of 65.14 pieces/m3. Geospatial visualizations revealed pollution hotspots near industrial coastlines and ocean current pathways. The study demonstrates a scalable solution for real-time pollution assessment and policy planning. Despite the high performance, limitations in real-time data and coverage of remote ocean regions persist. Future work will focus on integrating satellite imagery, IoT-based sensors, and real-time analytics to improve predictive capacity and support global marine pollution mitigation efforts.