Cost-effective mechanism for environmental toxic reduction using deep learning. (August 2022)

• Record Type: Journal Article
• Authors: Jun, He
• Is Part Of: Sustainable energy technologies and assessments. Volume 52:Part C(2022)
• Publication Date: 2022-08
• Subjects: Deep learning -- Environmental pollution -- Toxics -- Economic -- Chemicals; Renewable energy sources -- Periodicals; Energy development -- Technological innovations -- Periodicals; Electric power production -- Periodicals; Energy storage -- Periodicals; 333.79
• DOI: 10.1016/j.seta.2022.102206
• Languages: English
• ISSNs: 2213-1388

Abstract

Abstract: Environmental toxic reduction (ETR) is necessary to utilize cost-effective, standardized toxicity tests confined to acute reactions to high dosages when conducting environmental risk assessments on chemical goods and effluents inside countries. Chemicals and organisms found in the surroundings harmful to human health are environmental toxins. Physical elements that disturb biological processes and creatures that cause sickness are examples of dangerous chemicals and their compounds. Exposure to environmental contaminants has a slew of negative consequences. People, animals, and plants can be harmed by toxic waste that finds its way into the ground, rivers, or even the air. Heavy metals like mercury and lead stay in the environment for long periods of time and build up. When people or animals consume fish or other prey, they typically absorb these hazardous compounds. These expenses indicate the qualitative degradation of the natural environment caused by economic activity and are referred to as degradation costs. Deep Learning (DL) has an advantage over other toxicity prediction approaches since it builds a hierarchy of chemical characteristics. A further advantage of DL is that it inherently supports multitask learning, which means that one artificial neural network (ANN) may learn about all of a substance's harmful effects as well as other useful chemical properties. The ETR-DL pipeline has been created to use DL for toxicity prediction. It normalizes the process of chemical representations of molecules. That is followed by computing several chemical descriptors fed into machine learning algorithms. It builds models, tests them, and then assembles the best ones into ensembles. It makes predictions about new chemicals' toxicity. The responsiveness, precision, reliability, balanced consistency, and the area under the output response characteristics are some of the most commonly used metrics for evaluating classification model performance of 95.12%.