Development of a Smart IT Service Request Prioritization Algorithm Using Machine Learning and User Behavior Analytics Compared with Conventional IT Service Management Approaches

The rapid growth of digital services has increased the number, diversity, and operational complexity of information technology service requests submitted to enterprise service desks. Conventional prioritization methods, including first-in-first-out processing, manual analyst judgement, fixed impact–urgency matrices, and static rule-based scoring, frequently fail to consider semantic information contained in request descriptions, historical requester interactions, service dependencies, workload conditions, and the probability of service-level agreement violation. This paper develops a Smart IT Service Request Prioritization Algorithm that combines natural language processing, machine learning, operational user behavior analytics, business-impact assessment, service-dependency analysis, service-level agreement risk prediction, and queue-ageing controls. The proposed framework estimates request severity, predicts service-level agreement exposure, calculates a multidimensional priority score, and generates an explainable queue ranking while retaining human decision authority. A modular benchmark design uses 16,338 English-language support tickets for priority classification and 24,918 anonymized information technology incidents for service-level agreement prediction and queue-ranking evaluation. The text-based classification model obtained an accuracy of 69.37% and a macro-F1 score of 68.52%, compared with 39.84% accuracy and 35.38% macro-F1 for fixed rules. The combined operational-context and user-history model achieved a receiver operating characteristic area under the curve of 0.7805, compared with 0.5749 for a conventional impact–urgency baseline. The proposed ranking method achieved a mean normalized discounted cumulative gain at ten of 0.8400, compared with 0.7740 for static priority and 0.6007 for first-in-first-out processing. The results indicate that intelligent prioritization can improve service request classification, service-level agreement risk detection, and queue ordering. However, user-history variables should remain secondary operational indicators because their independent predictive value was limited.