Websites have become increasingly important in people's lives, fulfilling a wide range of needs across various domains such as shopping, education, news, and booking. Among the most heavily used website categories are online shopping platforms, whose usage has particularly increased during the COVID-19 pandemic, as they eliminate time and geographical barriers, providing access to a broader customer base. For these websites to effectively meet user needs and deliver a positive experience, they must be well-designed and adhere to usability principles. However, some existing shopping websites are poorly designed and do not follow usability best practices, resulting in suboptimal user experiences. Traditional manual website evaluation methods are time-consuming, and there is a need for more intelligent, automated approaches, particularly those leveraging machine learning techniques. This study aims to assist fashion shopping website developers in improving the usability of their platforms by providing an intelligent approach that can evaluate website usability. The study employs two complementary approaches for the evaluation process. The first model utilizes a Support Vector Machine (SVM) to assess websites based on specific usability principles, while the second model is a Convolutional Neural Network (CNN) that evaluates websites using features extracted from their screenshot images. The datasets for this project were custom-built, comprising a textual dataset for the SVM model and a screenshot dataset for the CNN model. The results demonstrate that the SVM model achieved an impressive 99% accuracy, while the CNN model attained 69% accuracy. These findings highlight the potential of this intelligent approach to provide comprehensive, data-driven insights for improving the usability of fashion shopping websites.

翻译：网站在人们生活中扮演着日益重要的角色，满足了购物、教育、新闻、预订等多个领域的广泛需求。在线购物平台是使用最频繁的网站类别之一，其在COVID-19疫情期间的使用量显著增长，因为它们打破了时间和地域限制，触达了更广泛的客户群体。为使这些网站能有效满足用户需求并提供良好体验，其设计必须完善并遵循可用性原则。然而，现有部分购物网站设计欠佳且未遵循可用性最佳实践，导致用户体验不佳。传统的人工网站评估方法耗时费力，亟需更智能的自动化方法，特别是利用机器学习技术的方案。本研究旨在通过提供可评估网站可用性的智能方法，协助时尚购物网站开发者提升平台可用性。研究采用两种互补的评估方法：首个模型利用支持向量机（SVM）基于特定可用性原则评估网站；第二个模型采用卷积神经网络（CNN）通过网站截图图像提取的特征进行评估。本项目数据集为自主构建，包含用于SVM模型的文本数据集和用于CNN模型的截图数据集。实验结果表明，SVM模型取得了99%的优异准确率，CNN模型则达到69%的准确率。这些发现凸显了该智能方法在提供全面数据驱动见解以改进时尚购物网站可用性方面的潜力。