Software development projects: A process of collecting and marking data

Abstract

During the recent years, software projects have tendency to be more complicated and sophisticated. Project Effort and timing preliminary estimation has low accuracy and IT-industry forces to use agile-methodologies to correct risks of resource overrun. Last inventions of software – neural networks and machine learning models in combination with classic methods like COCOMO offer help to estimate project’s parameters, but these models can solve only the tasks after continuous learning on big datasets.

One of the main problems in current project estimation research area is lack of available big highquality datasets of real software projects. All current existed datasets have low relevance, tight range of project features or are not available for wide network of researchers. This article describes new project’s
data mapping and collecting algorithm, divided to 3 different scenarios. it offers possibility to collect project data without complicated features or spending of additional resources. Next important part is idea of central
open-source repository for project’s data collecting. Implementation of this idea in project team and research community could seriously improve the quality of project’s estimation, accelerate continuous delivery of project results, improve research quality and extend research community.