INCIDENT MANAGEMENT

PROJECT AIM

A dashboard that provides actionable metrics for response and resolution times to enhance decision-making by encouraging tailored discussions through highlighting patterns and behaviours, enhance business reputation and answer daily questions in incident management environments such as:

What priority should be assigned to a ticket?

Who is responsible for this ticket?

This involved the use of agile methodologies to maintain industrial project management best practices, Python programming to evaluate, learn and manipulate the data for predictive modelling, implementation, etc. and finally, Power Bi to create visualisations and metrics that capture incident management concepts effectively.

THE INCONSISTENCIES

The department utilises DevOps to log, prioritise, and assign tickets to the relevant teams and members. While the department follows Information Technology Infrastructure Library (ITIL) standards for ticket processing, my experience in incident management has highlighted significant variations in how tickets are handled by IT Services team members with differing skill sets over time.

This inconsistency led to several issues:

  • Unstandardised prioritisation

  • Reliance on human intervention

  • Reoccurrence of similar incidents

SELECTING ALGORITHMS

To address these problems, live DevOps data was chosen for the dashboard, offering extensive insights into ticket activity. However, it was crucial to evaluate how the inconsistencies in processes might influence algorithm performance. Selecting suitable algorithms was essential, and since no single algorithm guarantees accuracy, the following were assessed:

  • Logistic Regression: Evaluates the probability of events or classes.

  • Multinomial Naïve Bayes: Predicts mutually exclusive events/classes using text classification.

  • Perceptron: An artificial neural network that predicts outcomes (Y) based on a single input (X).

I conducted an experiment to evaluate the performance of these algorithms using 70% of the data for training and 30% for testing. The best-performing model was selected and implemented in the dashboard project to enhance incident management processes.