ARTIFICIAL INTELLIGENCE APPLIED TO SOFTWARE ENGINEERING

Abstract

This work addresses the application of Artificial Intelligence (AI) in software engineering, highlighting how different techniques can contribute to improving development processes, reducing risks and estimating costs more accurately. The main objective of this production was to investigate and present the main AI methodologies used in software engineering, with a special focus on neural networks, fuzzy logic and Bayesian networks, analyzing their concepts, practical applications and results obtained in case studies. The methodology used was bibliographical research, based on works and scientific articles by various renowned authors in the field, such as Pfleeger, Pressman, Haykin, Barros and Cozman, among others. As a result, it was found that neural networks are effective in estimating software project costs and efforts, using historical data to improve the accuracy of forecasts. Fuzzy logic proved to be a suitable tool for assessing software quality, taking into account subjective and imprecise factors typical of real environments. Bayesian networks, in turn, stood out for their ability to model uncertainties and establish causal relationships between variables, providing useful simulations for decision-making. It can be concluded that the incorporation of these AI techniques into software engineering represents a significant evolution, promoting gains in efficiency, quality and reliability of the systems developed. Furthermore, there is a need for further research and practical applications in order to maximize the potential of these tools in the context of software development.