A Machine Learning Based Approach for Recommending Unfamiliar Process Activities

Abstract

Performance of business processes is affected by capabilities of employees who execute activities in these processes. Resource allocation approaches aim to identify an optimal allocation of resources to activities which supports the achievement of some process performance goals (e.g., in terms of time or cost). The majority of resource allocation approaches proposed in the Business Process Management community are process-centric (i.e., they aim to optimise process performance) and they neglect capability development needs of employees. In this article, we propose a novel approach for recommending unfamiliar process activities to employees which is based on the application of machine learning techniques to information extracted from process execution data. The goal of the approach is to assist organisations in their quest for capacity development by providing employees with opportunities to gain experience through the execution of new activities. The approach was implemented and evaluated by conducting experiments with real publicly available event logs. In the experiments, we compared the predictions provided by the approach with actual activity executions recorded in the logs. The experiments demonstrated the effectiveness of different approach configurations and showed that this machine-learning based approach significantly outperforms an existing algorithm proposed in earlier work.