基于KNN算法与局部回归的网站无障碍采样评估

提出一个新的抽样评估方法，通过对评估样本的KNN分析，选出特定网页.因大规模网站抽样结果稀疏，KNN算法会导致高检测误差，应用一个局部回归模型提升KNN评估质量.首先在网站中随机选择一些网页进行评估，得到该网站初始无障碍得分.在此基础上，将每一个评估网页作为一个标记样例，其他网页根据KNN局部回归模型进行无障碍评估得分预测.实验结果证明：所提方法相比随机抽样算法的效果上有着显著性提升.
Abstract: A novel sampling evaluation algorithm was proposed for a given page based on the KNN evaluated samples. As sampling in a large website tends to be sparse, KNN may lead to a high evaluation bias and a local regression model was thus employed to improve the quality of KNN-based evaluation. First, a certain number of webpages were randomly selected from a website and evaluated to obtain an initial website accessibility score. Each evaluated webpage was treated as a labeled sample and the accessibility scores for the rest pages in the website were estimated using local regression on the KNN. The experimental results validate that the proposed algorithm has significant improvement over the random sampling algorithm in website accessibility evaluation.