Abstract: In view of the complex and redundant problems of the evaluation index in the seafood safety assessment system, a new attribute unimportance algorithm is proposed based on mixed information granularity. The algorithm firstly analyzes the feasibility of attribute unimportance in theoretically, then it only calculates arbitrary core attribute in order to reduce the algorithm’s time complexity. Finally, it selects a small attribute unimportance value and determines the best set of attributes reduction by comparing the other attributes unimportance value. Through the experimental analysis and validation, it shows that the algorithm can be applied to reduce the assessment index in the seafood safety assessment system.

Key words:  attribute reduction, seafood safety, information granularity

摘要： 针对海产品安全评估体系中大量的、冗余的评估指标问题，提出了一种新的混合信息粒度的属性不重要性算法。该算法首先从理论上分析了属性不重要性的可行性，提出了属性不重要性公式，然后为了减少算法的时间复杂度，任意求出算法的一个核属性，最后，通过比较其他属性的属性不重要性值，选择属性不重要性值小的属性，确定最佳的约简属性集。通过具体实例分析和实验结果，表明该属性约简算法可以应用到海产品安全评估体系中。

关键词: 属性约简, 海产品安全, 信息粒度