Multi-criteria selection of the specialist for task execution in the software development process

Abstract

The object of this study is a process of selecting executors for tasks in software development among the available specialists. The problem being solved is the insufficient accuracy and explainability of existing approaches to selecting a specialist caused by the incompleteness of the characteristics of specialists, in particular soft skills. This can lead to delays in delivery, insufficient quality of delivery, and inefficient use of labor resources in the software development process. The article proposes an information technology for selecting an executor, which is based on a specialist model and a method for selecting an executor for a task. The specialist model takes into account the historical dynamics of possession of technical and social-communicative skills, the success of performing previous tasks, domain expertise, roles in projects, and the context of their execution. The key feature of the developed technology is the formalized representation of the specialist's profile in the form of a structured tuple with classifiers of competence and skill levels, which allows for a quantitative assessment of compliance with task requirements. The model includes seven main components: personal data, education, current role, role history, technical skills, domain knowledge, and soft skills. For each characteristic category, separate classifiers have been developed, which are converted into numerical values for calculations. Based on the created model, a method for evaluating the compliance of a labor resource to a task has been built, which takes into account the weight coefficients of requirements, normalizes the values of characteristics, and provides an explainable compliance metric. The method implements a four-step process: determining the characteristics necessary for task execution; representing specialists in terms of their compliance with task requirements; computing the compliance function with weight coefficients; searching for the best candidates among available specialists. The technology allows automating the process of selecting specialists, taking into account both the current state and the development of competencies over time. Experimental evaluation demonstrated a 41% reduction in decision-making time compared to the traditional approach. The results are explained by the use of a clear structure of characteristics, mathematical formalization of the compliance function, and validation under real operating conditions. The technology can be effectively used in organizations provided there are historical data on the activities of specialists that are guaranteed to be reliable, the dynamics of competence and skill development, and sufficient detail of requirements for the tasks being performed. The technology limitations are the dependence on the reliability of the input data and its inexpediency for use for tasks with an expected effort of less than one man-week.