面向不完备邻域系统的三支决策粒计算方法

针对许多应用领域中的数据大多是连续型数值，且由于成本限制和隐私保护等因素导致数据呈现不完备性。为此，提出面向不完备邻域系统的三支决策粒计算方法，引入邻域概念改进非对称相似关系，并通过改进的非对称相似关系计算不完备邻域系统的邻域粒度；在此基础上计算对象属于决策类的最大条件概率，再根据三支决策规则分别将对象划分到各决策类的正域、负域和边界域；通过实验比较和分析，进一步验证了该方法能获得较好的划分准确率和较低的误划分损失，为三支决策对不完备邻域数据的分析和处理提供了一种有效可行的方法。

Incomplete Neighborhood System Oriented Three-Way Decisions Granular Computing Method

Aiming at the continuous and incompleteness data that due to cost constraints and privacy protection in many applications, and appears, the incomplete neighborhood system oriented three-way decisions granular computing method is proposed. The asymmetric similarity is expanded through the neighborhood concept and neighborhood granularities of the incomplete neighborhood system based on the improved asymmetric similarity relations are calculated. The max conditional probability that the object belongs to the decision type is computed on this basis, and objects are classified into the positive, negative and boundary region of each category according to the three-way decision rules. Through experimental comparison and analysis, it is further verified that this method can obtain higher classification accuracy and lower misclassification loss, which provides an effective and feasible granular computing method for three-way decisions to analyze and process incomplete neighborhood systems.