Aprendizaje automático aplicado al desarrollo de software

Software development faces constant evolution due to the impact of new technologies, particularly machine learning, which has shown significant potential to improve efficiency and accuracy in various stages of the process. As the software industry continues to move towards greater automation, it is crucial to evaluate how these techniques are being adopted and what the main limitations are. This research aims to determine the current state of machine learning application in software development, synthesizing the methods used and formulating recommendations for future research. A systematic literature review was conducted using the PRISMA methodology, which ensured an accurate view of the available knowledge. The main results indicated a significant increase in the adoption of machine learning, especially in failure prediction, code optimization, and error detection, with the most commonly used techniques being Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and unsupervised clustering algorithms like K-means. However, limitations related to data quality and the need for high computational resources were identified. It is concluded that although machine learning is transforming software development, technical and ethical challenges -such as dependence on data and resources- must be addressed to maximize its impact. Therefore, it is recommended to continue researching approaches that optimize its implementation, ensuring fairness and transparency in the developed systems.

• Hlavní autor: Chora Paucar, Ricardo Javier (author)
• Médium: bachelorThesis
• Vydáno: 2025
• On-line přístup: https://dspace.ueb.edu.ec/handle/123456789/7919