Automating Accessibility Compliance: Leveraging Machine Learning to Analyze Developer Challenges with WCAG Guidelines

The prevailing strategy for ensuring web application accessibility relies heavily on adherence to established guidelines, particularly the Web Content Accessibility Guidelines (WCAG). However, the process of interpreting and implementing these guidelines presents significant challenges for developers, which are often compounded by the time-consuming nature of manual analysis required to ensure compliance. Despite the critical importance of accessibility, there remains limited exploration of these challenges within the development process. In this paper, we aim to address these challenges by leveraging machine learning techniques to automatically classify and analyze accessibility-related issues discussed by developers on Stack Overflow, with a focus on understanding their connection to WCAG 2.2 guidelines. To achieve this, we analyzed 5,644 developers' posts related to web accessibility collected from Stack Overflow between 2008 and 2022. These posts were categorized according to the principles of WCAG 2.2 using advanced machine learning models, including Random Forest, Support Vector Machine (SVM), Gradient Boosting, AdaBoost, and Multi-Layer Perceptron (MLP). The results highlight a strong emphasis among developers on addressing complex, non-linear accessibility issues. Our analysis demonstrated that Random Forest and Gradient Boosting models were most effective, achieving accuracy rates of 0.987 and 0.986, respectively. These findings underscore the challenges developers face in aligning with accessibility standards, particularly in areas such as time-based media customization and screen reader configuration, and the potential for machine learning to support more efficient and accurate compliance with accessibility guidelines.