An Integrated Artificial Intelligence-Based Model for Automatic Classification and Assignment of It Service Desk Tickets

Esquivel S.; Bruzza M.; Molina L. · International Conference on eDemocracy and eGovernment, ICEDEG (2026) · DOI: 10.1109/icedeg70169.2026.11695462

Resumen

Manual ticket management at the IT Service Desk of the Ministry of Economy and Finance (MEF) of Peru generates delays, misassignments, and limited traceability due to dependence on manual classification and routing processes. This study proposes and evaluates an integrated artificial intelligence-based model designed to automate critical stages of the service workflow. The proposed solution includes: (i) automatic ticket classification using natural language processing techniques, (ii) a chatbot with natural language understanding capabilities for initial support and request logging, and (iii) a rule-based weighted assignment engine that routes tickets according to priority, incident type, workload, and technical specialization. A dataset of 127,691 anonymized historical records (May 2022-August 2025) was used to train and evaluate a TF-IDF + SVM classifier. Due to the pronounced class imbalance, SMOTE resampling was applied during model training. The classifier achieved 94% overall accuracy and 92.4% average cross-validation accuracy (k = 5). In addition, the chatbot's NLU model achieved 98% precision and recall in intent classification and 94% success in controlled functional testing scenarios. Based on BPMN-driven process redesign and simulation, the proposed workflow is projected to reduce classification and assignment times by approximately 40% and operational workload by up to 50%, subject to confirmation through pilot deployment in a production environment. The findings suggest the potential feasibility and institutional applicability of AI-assisted automation for improving operational efficiency and traceability in public-sector IT service management environments.

Instituciones: Pontificia Universidad Católica del Perú