Anomaly Detection of Employee worklogs using Ensemble Learning

Abstract

"Being working in an organization where projects are carried out in line with agile practice, all resources log time to a task or a story that he or she is tagged to. Delivery managers, Program managers and Project Managers are responsible for governing and making sure all resources log time in an efficient manner to make sure that there are no major deviations of the initial scope and there is no deviation from the planned resource effort. But more often due to project scope volatility, client dependencies or project management malpractices the projects get dragged and deviate from initial planned delivery. In project audits we often find out that these could have been detected if we had scrutinized the time sheet and had a proper mechanism to predict the anomalies whereas BI team, we could have taken proactive actions. On a monthly basis we have ~35000-worklogs accumulated and analysing this along with the planned project allocation, initial effort estimation, project completion to find anomalies is manually a very exhausting effort which currently done manually during practice review. This gives a need for a tool to identify anomalies beforehand using the already identified anomaly patterns. As an organization the main objective is to identify time logs with anomalous timesheet entries for further scrutiny such that in weekly reviews our delivery managers can proactively look at these behaviours and get feedback from Project managers proactively rather than reactively and drill down on the anomalies which may reveal some ground level problems or dependencies that the ground level team is facing or encountering. "