Ticket Classification Using Machine Learning

Incident management is one of the key functions of IT Service Management and plays a major role in ensuring quality and efficient services for the enterprise users. Primary focus of the IT helpdesk is to provide first-level solutions, but in many cases, tickets are transferred to subject matter experts for deep dive analysis and issue resolution. The transfer and assignment processes followed are still manual and many a times result in human errors, resource consumption, higher response and resolution times, and ultimately poor customer service. Introduction of automation in incident classification, tagging and assignment cannot only reduce issue resolution times but also improve customer satisfaction. The main purpose of this research is to focus on automated approach for analysis and classification of “incidents” using machine learning (ML) or natural language processing (NLP) techniques. IT helpdesk ticket classifier models can be trained by mining the historical ticket inventory datasets. NLP techniques can be useful in extracting keywords from incident description while ML-based algorithms to classify and tag the incident. For ticket classification, we used six ML classifiers, and the experimental results show that logistic regression, decision Tree, and SVM models were found to have better accuracy and F1 score as compared to other considered classifiers.