Building Intelligent Service Desk Systems using AI
S. P. Paramesh, K. S. Shreedhara · 2019
Service desk system helps the users of the organization to raise the problem tickets by selecting the appropriate problem category and to obtain the solution to user problems. Manual selection of ticket category may result in dispatching the tickets to incorrect resolution group. Incorrect assignment of service desk tickets in turn results in reassignment of tickets, resolution time delay and greatly affects the business. The current service desk system does not handle the user’s unstructured ticket data efficiently. So artificial intelligence concepts like machine learning and natural language processing can be applied to automate the existing service desk systems. In this research paper, we developed an automated service desk ticket classifier model that automatically categorizes the incoming ticket by analyzing the unstructured natural language ticket description entered by the end user. Supervised machine learning techniques like Decision trees and Naive Bayes models are used to build the service desk ticket classifier system. Ensemble models are also used to analyze the performance of ticket classifier. Ensemble of decision trees, naive bayes and Random Forest classifiers are used as a part of ensemble classifier techniques. The performance of service desk ticket classifier models is evaluated on a real world IT infrastructure ticket dataset and compared using various performance metrics. Ensemble based Random Forest classifier performed well when compared to all other considered models. The proposed model for the auto categorization service desk tickets results in simplified user interface, faster ticket resolution, efficient resource utilization and improved growth in business.