AUTOMATIC TUNING WIZARD FOR FT STAFF ROSTERER

Abstract

FT Staff Rosterer is a customizable workforce management system which can be integrated with the client's existing IT infrastructure. The system provides wide variety of modules for staff to concurrently register their preferences, duty planner to plan and generate the roster, and a totally integrated messaging system that allows staff to communicate shift change and swap request amongst themselves and duty planners, with the aid of handheld devices like PDAs, cell phones and pagers. Just as any other software, it is crucial for the user to configure their Rosterer properly in order to generate the desirable results. However, the configuration of FT Staff Rosterer has proved to be a highly non-trivial task and consultation from experts is inevitable during the deployment process. Moreover, the customer's needs are constantly changing in the current environment of business and they require that the software should be easily reconfigured without the resort to another deployment consultation. The automatic tuning wizard for FT Staff Rosterer attempts to tackle this problem from a machine learning approach. By using this wizard, the users are kept away from directly adjusting the optimization parameters, whose number sometimes can count up to hundreds. Instead, they merely need to provide the system with samples of desirable and undesirable rosters. Consequently, the system will try to learn all the optimization parameters from these samples. Through this approach, the deployment and re-configuration process could be made easier and cheaper, which increases the productivity of both the company and the customers.