Author(s)

Jin Zhou, Richard Irampaye, Xu E, Yunfeng Liu and Yanhong Li

Corresponding Author

Xu E

ABSTRACT

In the real-world, some decision systems experience a dynamic variation of their attributes and attribute values over time. With this age of sensor technology and the Internet of Things (IoT), multidimensional data hard to process is generated. Some tools based on partitioning into small subsets relatively easy to process have been developed for the multidimensional data. However, they are statics and cannot be used in a changing environment. The attribute reduction is critical to handle. In this paper, we mainly focus on the update of the attribute reduction along with the time complexity improvement. The multidimensional data is divided into subsets, and then we compute the core for each subset and the global core is the union of all subsets’ core in the system, after that the attribute reduction is computed. The new data type enters into the system as new subsets to fuse with the existing subsets. Any information for update will be processed on subset level. We may have scenarios such as adding, removing, adding and removing simultaneously along with the new data type to enter into the system. The updating process results in positive region change, which imply a computation of a new core and new attribute reduction, which will be done dynamically over time. We use the discernibility matrix method for computation. This algorithm avoids some re-computations by using existing subsets and core for those with the unchanged positive region. Some examples provided to illustrate the proposed algorithm.

KEYWORDS

Rough set; information update; multidimensional data; core attribute; partitioning; attribute reduction