Enhancing Safety Data Analysis Through Machine Learning Models

Abstract Currently, there is a positive trend in the oil and gas industry worldwide, including Kazakhstan, in the directions of digitizing processes and ensuring occupational safety. As a result, we have a large volume of the data, the potential of which is often not fully realized. The second derivative of improvements in occupational safety is the exponential increase in task complexity as we approach the coveted goal of incidents. In other words, achieving each subsequent safety metric requires making increasingly complex decisions. We observe how two curves—the demand curve and the opportunities curve—intersect at this point, which is now. The main goal of the study is to develop a solution that unlocks the potential of the collected data and applies it to make those very complex decisions. This can be achieved by uncovering non-obvious, complex interdependencies, causal relationships, and behavioral patterns. The research focuses on centralizing safety related data to ensure its accuracy for predictive, prescriptive, and diagnostic analysis. The main goal is to anticipate and predict incidents by leveraging historical incident data along with regular observations and planned works. Aggregating information from various sources such as incident reports, work permits, and safeguards field verification is essential for constructing a holistic view of safety outcomes. By combining these datasets, we gain a deeper understanding of the factors influencing safety and risk management.