Abstract: Evaluation data is easily polluted by noise data, so conclusion of evaluation is usually distorted. A new noise cleaning arithmetic is provided to find out and filter noise data. Imitating experiment proves that this arithmetic can find out and filter the noise. This arithmetic is useful in the field of data management.

Key words: evaluation data, noise cleaning, arithmetic, imitating experiment

摘要： 评价数据易被噪声污染，导致评价结论失真。该文提出一种基于数据密度的噪声清洗(DNC)算法，用于识别并过滤评价数据中的噪声，介绍了一套仿真实验方案。实验结果表明，DNC算法可以提高噪声的识别和过滤效果。该算法在数据管理领域具有应用价值。

关键词: 评价数据, 噪声清洗, 算法, 仿真实验

CLC Number: 

• TP301.6