Effort Estimation using Bayesian Networks for Agile Development

Abstract

This article proposes an automatic method to estimate the effort of development based on narrative texts. Narrative used the agile method. We propose techniques of symbolic analysis of natural language were used for extraction of verbs and nouns, and verbal reduction (verbs in infinitive), and the standardization of keywords through synonyms. For the machine learning was used if the naive Bayesian classifier model. Apply and test the model in real environment that used the narratives in the Portuguese language in the form of BDD (Behavior Driven Development). In these tests, obtaining an accuracy of 83% in the estimates of the story points.