广义多尺度决策系统的局部最优粒度选择

在实际应用中,人们常常选择比较合适的粒度层次来解决相应的问题。在经典的多尺度决策系统和粒度层次构造过程中,属性取值常由人工选择某些固定粒度层次。本文针对广义多尺度决策系统,由属性取值的尺度组合来构造粒度层次,进而研究局部最优粒度的选择问题。首先,介绍了广义多尺度决策系统的概念。然后,在协调的广义多尺度决策系统中定义了最优粒度和局部最优粒度,并给出了基于属性组合的最优粒度与局部最优粒度的选择算法。最后,在不协调的广义多尺度决策系统中引入了广义决策,定义了广义决策最优粒度和广义决策局部最优粒度,并给出了基于广义决策最优粒度与广义决策局部最优粒度选择算法。

Local optimal granularity selections in generalized multi-scale decision systems

People tend to choose the appropriate level of granularity to solve problems in application. In the classical multi-scale decision systems, during the process of constructing the levels of granularity, some fixed levels of granularity are selected manually for attribute values. Aiming at generalized multi-scale decision systems in this paper, the levels of granularity are constructed by using the scale combination of attribute values. Furthermore, the selection problems of local optimal granularity are studied. The concept of generalized multi-scale decision systems is introduced firstly. Then, notions of the optimal granularity and the local optimal granularity in consistent generalized multi-scale decision system are defined, and the algorithms for finding optimal granularities and the local optimal granularities are described. Finally, the generalized decisions are introduced to inconsistent generalized multi-granular decision systems, the generalized optimal granularity and the generalized local optimal granularity are defined, and the algorithms for finding the generalized optimal granularity and the generalized local optimal granularity are also investigated.