Machine learning techniques as an advanced hybrid framework for improving crisis management in investment sectors: an experimental study

Abstract

Risk management is a fundamental element in today’s vital investment sectors due to the growing scale and complexity of economic data. Evaluating business risks, plans, and timelines using traditional methods—including analytical, technical, and inferential models—has become insufficient for addressing the complex structures of modern datasets.

This study aims to develop a comparative hybrid framework that integrates machine learning techniques to enhance crisis management and reduce its impact on critical sectors such as financial investments, oil and gas fields, and real estate. The proposed framework employs linear regression, classification, and clustering, with a focus on the bias–variance trade-off to ensure an optimal balance between accuracy and generalization.

The study demonstrates how these techniques can be used to predict risks, detect patterns, and support strategic decision-making processes (before, during, and after crises) to mitigate crisis severity. The framework utilizes several evaluation metrics: classification performance is measured through accuracy, precision, and recall; regression performance is assessed using the root mean square error (RMSE); and clustering quality is evaluated using the Silhouette Score. Together, these metrics provide a clear and balanced view of model performance and generalization capabilities.