PREDICTING INCIDENT RESOLUTION TIME IN DIGITAL TRANSFORMATION SYSTEMS AND ITS IMPACT ON DECISION-MAKING USING MACHINE LEARNING

Abstract

The study aims to develop a predictive model. It depends on techniques to learn how to predict incident resolution times in digital transformation systems, and to improve practical decision-making, and achieve more efficient incident management in Information Technology environments. The dataset contains 141,712 incident records, it was completely collected from the platform ServiceNow, with 36 variables describing different aspects of incident management. This research applied several machine learning models, such as Bayes Net, Random Forest, and Support Vector Machine, where we showed that the Support Vector Machine model achieves a higher accuracy of 95%, while the Random Forest showed balanced performance, with an F1 Score of 75%. The challenges associated with predicting incident resolution time were analyzed, including the inaccuracy of traditional models and their reliance on unlabeled data. This study makes a scientific contribution by developing an accurate machine learning-based model that contributes to improving the efficiency of incident management and decision-making in digital transformation environments.