A Mathematical Modeling and Explainable AI for Intelligent IT Service Priority Optimization Using Accelerated XGBoost

Abstract

Impactful prioritization of incidents is a vital ITSM process that has a direct impact on service availability, operational costs and SLA fulfilment. This paper introduces a mathematical modelling framework with explainable artificial intelligence (XAI) for intelligent IT service priority optimization. In reality, the rule-based and manual prioritization systems are not scalable, and they do not consider complex interactions between attributes of incidents that may lead to delayed response and possible misclassification of critical incidents. To overcome these problems, we propose an explainable and high-accuracy machine learning framework for predicting priorities of ITSM incidents using mixed-type ITSM data (numerical and categorical features) and applying structured feature engineering and ITSM data preprocessing for these data types. The framework is tested in a large scale real world ITSM dataset that has a strongly imbalanced class distribution. The model successfully achieves a training accuracy of 99.99%, a testing accuracy of 99.99%, a macro F1 score of 1.00 and perfect precision and recall in detecting high-priority incidents with only a slight gap between training and testing suggesting that the model is not merely memorizing the training set. The most influential features for priority determination are identified in SHAP analysis as urgency level, impact score, reassignment count and incident resolution time.