Cardiac Risk Screening for Multimodal Monitoring and Intelligent Prediction of Unsafe Behaviors in Oil and Gas Workers ——An RFE-Optimized RF-MLP Model

Huang Chuanhong^*, Zhang Xin, Chen Yanhong, Zhang Yuqi

College of Safety Engineering, Chongqing University of Science and Technology

Abstract：

Unsafe behavior in oil and gas operations may arise from interactions among physiological load, psychological state, environmental stressors, and task conditions. Acute cardiac dysfunction is especially important because it can abruptly impair attention, motor control, and decision-making, thereby increasing the likelihood of secondary accidents. As an initial physiological-risk module within a planned multimodal monitoring framework, this study develops an RFE-optimized random forest and multilayer perceptron (RF-MLP) model for cardiac risk screening. The publicly available UCI heart disease dataset was used for algorithm development. Random forest-based recursive feature elimination identified a compact subset of six clinically meaningful variables, and the MLP learned nonlinear associations between these variables and the heart-disease label. In the reported experiment, the model achieved an accuracy of 0.920, an AUC of 0.980, a recall of 0.905, a precision of 0.912, and an F1 score of 0.908, outperforming the comparison models. The selected variables provide a concise representation of cardiac reserve, myocardial ischemia, and baseline physiological vulnerability. These findings support the feasibility of using feature-screened cardiac data as one input to a broader physiological-behavioral-environmental-psychological prediction system. Because the dataset is clinical rather than occupational and contains no observed unsafe-behavior outcome, the present results should be interpreted as proof of concept for physiological risk screening, not as validation of unsafe-behavior prediction in oil and gas workers. Field data and external validation are required before operational use.

Key Words：

multimodal monitoring; oil and gas workers; unsafe behavior; cardiac risk; RFE; RF-MLP