We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Integration of bio-inspired algorithms with machine learning for modeling microplastic contamination transport processes in natural aquatic habitats using morphological features
Summary
Scientists built a computer model that can accurately predict how microplastic particles move through rivers, lakes, and oceans based on their shape and size. This matters because understanding where these tiny plastic bits travel and accumulate, including in water sources we may eventually drink or in seafood we eat, can help authorities track pollution hotspots and create smarter policies to keep our water supplies cleaner.
The increasing presence of microplastics (MPs) in aquatic environments poses significant environmental challenges, necessitating accurate prediction models for their transport behavior. This study develops a machine learning (ML)-based predictive framework to estimate MPs' speeds using artificial neural network (ANN) integrated with optimization techniques. Key hydrodynamic and particle properties influencing MP transport were identified using the Minimum Redundancy Maximum Relevance (MRMR) algorithm. Also, local interpretable model-agnostic explanation (LIME) was used in the current study to show the individual contributions of each feature. The comparative analysis of many optimization techniques, such as Particle Swarm Optimization (PSO), Firefly Algorithm (FA), Artificial Bee Colony (ABC), Bayesian Optimization (BO), and Random Search (RS), was implemented to optimize the ANN. The obtained results indicated that ANN-ABC-C1 (r = 0.720, RMSE = 0.037, MSE = 0.0006) and ANN-ABC-C2 (r = 0.764, RMSE = 0.035, MSE = 0.00048) display poor predictive ability, and thus it can be concluded that ABC is not particularly well-adapted to optimizing ANN models in this context. Conversely, Bayesian BO and RS perform better. ANN-BO-C2 (r = 0.971, RMSE = 0.019, MSE = 4.87E-05) and ANN-RS-C2 (r = 0.981, RMSE = 0.017, MSE = 2.46E-05) are the best models, which proves that they are highly generalized. ANN-RS-C2 model has the highest predictive accuracy, with r (0.981) and the RMSE = 0.017, and MSE = 2.46E-05. The research equally highlighted the role of modeling in environmental pollution analysis, whereby MP dispersion and accumulation are predicted. The findings are useful in terms of pollution prevention, the development of regulatory policies, and sustainable planning. The proposed framework offers a solution to the problem of real-time MP transport forecast that will contribute to the sound management of water resources and the protection of the ecosystem.