A Novel Multi-Objective Evolutionary Algorithm to Address Turnover in the Software Project Scheduling Problem Based on Best Fit Skills Criterion

Abstract

As a typical scheduling problem, the software project scheduling (SPS) under uncertain and dynamic events has gathered attention of academic researchers and practitioners because it causes tangible and intangible loss and will also influence the project schedule and cost if not addressed in time. In this paper, a SPS model based on skills acquired by employees and required by tasks criterion is proposed to tackle the issue of ‘employee turnover’ along with novel multi-objective evolutionary optimization algorithm to handle such dynamic events by employing the domain knowledge for population initialization. This is critical for the success of a software project as when an employee leaves, the project is rescheduled with remaining resources to assess if the project can be completed on time and under budget; otherwise, an alternative resource is hired. The proposed model and algorithm are evaluated on 18 dynamic benchmark and 6 real-world problem instances. The experimental results indicate that if existing employees are reassigned to the tasks, the project cost and duration increases on average 22% and 4% respectively. However, hiring new employees with recommended skills and proficiency based on proposed approach provides 9% decrease in project cost with no increase in duration and 15% decrease in cost against 11% referenced increase in project duration, in comparison to classical rescheduling results. The results demonstrate that proposed model and algorithm handles dynamic event effectively and help in identifying the right set of skills for hiring of an alternate resource while optimizing project duration and cost.

Keywords

• Software project scheduling
• employee turnover
• meta-heuristics
• multi-objective optimization