The growing complexity and number of software projects requires both increasingly more experienced developers, testers and other specialists as well as a larger number of persons to fill these roles. This leads to increased personnel and management costs and also makes effort and cost estimation at task and activity levels more difficult for software development companies. An automated solution for software development effort estimation based on text descriptions of tasks and activities, combined with available metrics, is introduced. A real world case study consisting of data from a software company whose activity spans a rich development spectrum is conducted. The results obtained are very encouraging and surpass the few similar approaches available in research literature.
Software development effort estimation, Machine learning, Word embeddings, doc2vec, Support vector machines, Gaussian Naive Bayes.
Vlad-Sebastian IONESCU, Horia DEMIAN, Istvan-Gergely CZIBULA, "Natural Language Processing and Machine Learning Methods for Software Development Effort Estimation", Studies in Informatics and Control, ISSN 1220-1766, vol. 26(2), pp. 219-228, 2017. https://doi.org/10.24846/v26i2y201710
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).