EnvHealthNet: A Multi-Modal Machine Learning Model for Commercial Environmental Health Risk Prediction

Abstract

The combination of pollution with industrial operations and urbanization creates major health problems for public health in commercial areas. The research presents EnvHealthNet as a Multi-Modal Machine Learning Model which predicts environmental health risks through the combination of environmental sensor information together with commercial activity records environmental and health data. The model combines Convolutional Neural Networks and Long Short-Term Memory networks and LightGBM along with advanced data fusion procedures to process its data inputs. The performance assessment shows that using multi-modal fusion exceeds single-model performances. EnvHealthNet reached exceptional 97.89% performance accuracy while exhibiting 98.0% ROC-AUC which surpassed traditional models such as CNN (85.2%), LSTM (86.7%) and LightGBM (88.1%). The identified risk factors that proved important during feature importance analysis included PM2.5 concentrations together with industrial zone proximity and noise pollution levels. Through the model scientists discovered high-risk locations inside industrial areas in addition to areas with low environmental risks that receive proper management. Environmental health risk predictions achieve a substantial improvement in accuracy and reliability when machine learning approaches adopt multi-modal techniques thus offering decision-making opportunities to commercial enterprises and policymaking bodies as well as environmental agencies that work to reduce health threats.