On the Viability of Using SRGMs for IT Help Desk Incident Predictions

In this case study we investigate software reliability models and their applicability to process improvement at an IT help desk. We propose a model selection framework and demonstrate its success using real help desk incident data from a portfolio of 156 desktop software applications. Incidents are predicted at five intervals and measured against actual numbers of submitted incidents. We analyze incident prediction accuracy, and report on the trend in accuracy based how far into the future incidents are predicted. Our results demonstrate a model selection framework can assist with predicting for a large portfolio of products. The level of accuracy reported in this industry-based case study establishes the proposed estimation technique and reliability model selection framework as novel research in software engineering. Additionally, its practical uses are applicable to help desk process improvement efforts.