An incremental attribute reduction approach based on knowledge granularity under the attribute generalization

Abstract

Attribute reduction is a key step to discover interesting pattern in decision system with numbers of attributes available. Moreover, data processing tools have been developed rapidly in recent years, and then the information system may increase quickly in attributes with time in real-life applications. How to update attribute reduction efficiently under the attribute generalization becomes an important task in knowledge discovery related tasks. The attribute reduction of information system may alter with the increase of attributes. This paper aims for investigation of incremental attribute reduction algorithm based on knowledge granularity in information system on the variation of attribute set. Incremental mechanisms to calculate the new knowledge granularity are first introduced. Then, the corresponding incremental algorithms are presented for attribute reduction based on the calculated knowledge granularity when multiple attributes are added to a decision table. Finally, experiments performed on UCI data sets and the complexity analysis show that the proposed incremental methods are effective and efficient to update attribute reduction with the increase of attributes.

Valid Incremental Attribute Reduction Algorithm Based on Attribute Generalization for an Incomplete Information System

“…Matrix servers are a type of important mathematical tool, and recently, they have been widely used in research on rough set theory [18,23,26,[35][36][37][38] . For example, Liu et al utilized a matrix to express the upper and lower approximations of a classic rough set [36] , and Jing et al used a matrix to represent the related knowledge granularity and perform attribute reduction [26] . Huang et al also applied a matrix to denote the approximate computation of a rough fuzzy set [35] .…”

“…The relation matrix has many advantages in the process of computing conditional entropy; however, the adoption of the matrix method causes the efficiency of the incremental updating to be low when the scale of the datasets is large, and the memory consumption is greater [26] . This means that when the complete information system attributes increase, consideration of the incremental computing through the view of the equivalence class changes of object [24][25][26] . In this section, we adopt this idea to incrementally update the conditional entropy based on the tolerance class of objects.…”

“…Li et al developed an algorithm of incremental update approximation for an information system that satisfies the dominance relation [25] . Jing et al researched two incremental update methods based on knowledge granularity, and they proposed two incremental attribute reduction algorithms for when the attribute is increasing [26] . However, these algorithms can only handle a complete information system, while in reality, some dynamic changes exist in information systems, such as the situation that attributes value is missing, which is called an incomplete information system.…”

“…However, for an incomplete information system in which the attributes increase, conditional entropy is seldom used to construct the incremental attribute reduction algorithm, and thus, we construct the incremental reduction algorithm using conditional entropy. A matrix is a common data analysis tool used in rough set theory [18,23,26,[35][36][37][38][39] . In this paper, we apply the matrix method to represent conditional entropy and use an incremental update mechanism of conditional entropy based on the matrix.…”

Attribute reduction of relative knowledge granularity in intuitionistic fuzzy ordered decision table

“…Further extended attribute reduction model. In the context of generalized attribute concepts, Jing and Li et al [35] extended the model of attribute reduction from knowledge granularity angles. In order to overcome that it is hard to update reduct when the large-scale data vary dynamically.…”