Integrative analysis and predictive modelling of IT service management data: A data-driven approach to informed decision making

Abstract

This thesis explores the integration of data-driven decision making (DDDM) in IT Service Management (ITSM), with a focus on incident management for a case company’s cloud managed services project. As a third-party provider, the company delivers cloud infrastructure, application management, and support services to its clients. Incidents are unexpected errors or failures in IT systems, and effective incident management ensures swift resolution to maintain seamless IT operations (Gupta, Mohania, & Prasad, 2008). The study employs descriptive and inferential analytics, time-series predictive modelling, and feasibility studies to gain more insight into the project, forecast daily ticket volume, and identify automation opportunities.

Using a dataset of over 31,000 tickets, the study uncovers 7-day seasonal patterns in ticket volume, statistically significant differences in business time to resolution (BTTR) - the time from ticket reporting to closure, calculated on business hours -across categories, and the category with the longest BTTR. Predictive models, including SARIMA and recently developed models such as Prophet, NeuralProphet, and Orbit’s Local Global Trend (LGL) model, demonstrate limited accuracy in ticket volume forecasting due to high mean absolute errors. Automation candidates are assessed based on both technical and economic feasibility. Key categories such as CPU, Disk Space/IO, Memory Usage, Storage, and Backup are identified as prime candidates for automation, representing nearly 48% of the total ticket volume. By automating these areas, significant manual intervention can be reduced at a minimal cost.

While predictive modelling faces challenges, this research provides actionable insights into leveraging analytics and automation to enhance ITSM workflows, improve service quality, and achieve cost savings. Recommendations include broadening the research scope, improving datasets for better prediction, and standardizing metrics across IT providers for cross-organizational analysis.