Font Size: a A A

A Comparison of Univariate Modeling and Forecasting Techniques for Predicting ITIL Demand Management Workload up to Six Weeks Ahea

Posted on:2018-03-12Degree:D.EngrType:Dissertation
University:The George Washington UniversityCandidate:Clavon, Dock Marshall, JrFull Text:PDF
GTID:1448390002999564Subject:Industrial Engineering
Abstract/Summary:
The City of Atlanta is actively pursuing tools and techniques to aggressively progress their ITIL Maturity level to achieve world-class performance in IT service delivery. Currently, a methodology to accurately forecast demand management to predict future workload does not exist. This added capability will ensure more accurate staffing forecasts against service targets. IT Infrastructure Library (ITIL) has proven itself to be a reliable framework to integrate into an Information Technology Service Management (ITSM) function. Although ITIL is well-known worldwide, little academic research has been published to date about the impact of low priority service requests impact on IT service delivery. This praxis presents the applied nature of modeling univariate IT demand management data.;The data collected in terms of date submitted, calendar week number of submission, resolution date, business hours to resolve, City of Atlanta department, issue type, and priority level were collected from January 2016 to December 2016. Using the Pareto Principle, one department was selected for the study with the two additional departments for purposes of validation, totaling 11,760 helpdesk tickets. The key metric, "Business Hours to Resolve," a mean-time-to-resolve (MTTR) metric, was aggregated into weekly averages totaling 53 discrete observations representing the 53 weeks in 2016 for each department. After necessary differencing, a model was selected that had the best characteristics with that of a properly-specified model. The models were validated by performing six-week ahead forecasts, and the predicted MTTR results were compared with the actual MTTR. A comparison of seven sophisticated univariate methods (1) Random walk; (2) Random walk with drift; (3) Constant mean; (4) Linear trend (5) Simple moving average; (6) Exponential smoothing; and (7) ARIMA was performed. The results indicate that the ARIMA model significantly outperformed the six other univariate methods and that the proposed methodology is suitable for developing future predictive ITIL Demand Management models for the City of Atlanta.
Keywords/Search Tags:ITIL, Demand management, Model, Univariate, City, Atlanta
Related items