Proposal for a predictive analysis of failure risk in software test cases using agile methodologies

This article presents a predictive model designed to estimate the probability of failure in the design of software test cases, based on contextual characteristics. By analyzing attributes such as the type of test, its origin, complexity level, and execution phase, the proposal aims to identify patterns that correlate with test failures. A synthetic dataset was generated to simulate a real agile development environment, allowing for the training and evaluation of a supervised machine learning model based on a neural network architecture. Subsequently, the model was compared with other classical approaches. The results highlight the potential viability of predicting failure risk from the design of test cases, which constitutes a methodological contribution to quality assurance strategies in agile software development contexts.