Modeling Employee Flexible Work Scheduling As A Classification Problem.

Area of Science:

• Computer Science
• Artificial Intelligence
• Operations Research

Background:

• The COVID-19 pandemic necessitated flexible working arrangements, exposing limitations in existing employee scheduling methods.
• Current scheduling algorithms are primarily designed for compressed work and struggle with the high dimensionality of constraints in flexible work environments.
• Automating employee scheduling with numerous constraints is a significant challenge for organizations.

Purpose of the Study:

• To propose a novel machine learning approach for automating employee scheduling in flexible working arrangements.
• To address the challenge of high-dimensional constraints in flexible work by reframing scheduling as a classification problem.
• To extract and utilize employee availability and user-defined constraints for improved schedule generation.

Main Methods:

• Developed a machine learning approach leveraging mining of user-defined (soft) constraints.
• Transformed the employee scheduling problem into a multi-label classification task.
• Automated extraction of employee personal schedules (e.g., calendars) to determine availability.

Main Results:

• The proposed machine learning approach achieved 93.1% accuracy in satisfying constraints for faculty staff scheduling.
• Results were comparable to a common constraint satisfaction and optimization method, which achieved 92.7% accuracy.
• Demonstrated the efficacy of using extracted employee availability and constraints in a multi-label classification model.

Conclusions:

• Machine learning offers a viable and accurate alternative for flexible employee scheduling, particularly with complex constraints.
• Automating constraint extraction and availability analysis enhances the efficiency and accuracy of scheduling systems.
• The developed approach provides a scalable solution for universities and other organizations adopting flexible work policies.