Yue-Jin Lv’s research while affiliated with Guangxi University and other places

After analyzing the attribute reduction algorithm based on Rough Set that has arisen at present, a new formula for measuring the importance of attribution was given, and the property of this formula was analyzed. Then a new algorithm for attribution reduction was provided. The time complexity of the provided algorithm is O(|C|2*|U|). At last, the efficiency of the new algorithm was illustrated with an example.

The concept lattice is useful in knowledge processing and analyzing. And it has been used with a high intensity to knowledge reduction and data mining. This paper, from the viewpoint of concept extents, studies new and relatively reasonable formulas measuring attribute significance and proposes a theory for justifying whether an attribute set is a reduction on concept lattice, and then uses those formulas as heuristic information to design a novel and heuristic algorithm for attribute reduction on concept lattice. Finally, a real example is used to demonstrate both its feasibility and effectiveness.

It is very time-consuming to discover association rules from the mass of data, but not all the rules are interesting to the user, a lot of irrelevant information to the user's requirements may be generated when traditional mining methods are applied. In addition, most of the existing algorithms are for discovering one-dimensional association rules. Therefore, this paper defines a mining language which allows users to specify items of interest to the association rules, as well as the criteria (for example, support, confidence, etc.), and proposes a method based on rough set theory for multi-dimensional association rule mining methods, dynamically generate frequent item sets and multi-dimensional association rules, which can reduce the search space to generate frequent itemsets. Finally, an example is used to illustrate the algorithm and verify its feasibility and effectiveness.

Based on the advantages of concept lattices and neural networks, this paper presents a concept lattice-based neural network model. In terms of Concept Lattice theory, we carry on attribute reduction of concept lattice and then the key elements are extracted, which can be used as input of BP neural network. Furthermore, the concept lattice-based neural network model can be set up after the sample training of BP neural network. Finally, the results of simulative experiment show that the model can simplify the BP neural network training sample, optimize the BP neural network structure and also enhance the study efficiency and precision of the system. So, this novel method is effective and feasible, what’s more, the theoretical significance and practical value are also outstanding.

Rough set theory is a new mathematical tool to deal with vagueness and uncertainty. And now it has been widely applied in constructing decision tree which has no hierarchical attributes inside. However, hierarchical attributes exist generally in realistic environment, which leads that decision making has max rules. Using max rules to build decision trees can optimize decision trees and has practical values as well. So, in order to deal with hierarchical attributes in decision tree, this paper try to design an optimize algorithm of decision tree based on rough sets hierarchical attributes (ARSHA), which works by combining the hierarchical attribute values and deleting the associated objects when max rules exist in decision table. So that the algorithm developed in this paper can abstract the simplest rule set that can cover all information for decision making. Finally, a real example is used to demonstrate its feasibility and efficiency.

The paper studies the fuzziness measure in fuzzy rough sets. By making use of the support set of fuzzy sets, a rough membership function for fuzzy sets based on fuzzy relation is introduced. Simultaneously, a fuzziness measure of fuzzy rough sets from total mean fuzzy degree is proposed. And then, it is proved that the fuzziness measure of fuzzy rough sets, denoted by, equals to zero if the set is crisp and definable.

This paper puts forward a new method of discretizing continuous attributes. Compared with the traditional approach, the method, proposed in this paper, can make the number of the obtained classes be more moderate, as well as the lost information be fewer. And then a simple attribute reduction algorithm is developed in continuous information systems. Finally, a real example is used to illustrate its feasibility and effectiveness, respectively.

Reduction of attribute is one of the key problems in rough set theory. In this paper, Using recursive idea, we design a new approach to calculate partition U/C, whose time complexity is O(C \U). Then two new and relatively reasonable formulas measuring attribute significance are designed for reducing searching space, which are used as heuristic information to develop a quick attribute reduction algorithm; the theoretical analysis shows that this algorithm is much less time complexity than those existed algorithms. Finally, experimental results demonstrate its feasibility and effectiveness, respectively.

Quantum Genetic Algorithm (QGA) is a promising area in the field of computational intelligence nowadays. Although some genetic algorithms to find minimal reduct of attributes have been proposed, most of them have some defects. On the other hand, quantum genetic algorithm has some advantages, such as strong parallelism, rapid good search capability, and small population size. In this paper, we propose a QGA to find minimal reduct based on distinction table. The algorithm can obtain the best solution with one chromosome in a short time. It is testified by two experiments that our algorithm improves the GA from four points of view: population size, parallelism, computing time and search capability.

Quantum Genetic Algorithm (QGA) is a promising area in the field of computational intelligence nowadays. Although some genetic algorithms to find minimal reduct of attributes have been proposed, most of them have some defects. On the other hand, quantum genetic algorithm has some advantages, such as strong parallelism, rapid good search capability, and small population size. In this paper, we propose a QGA to find minimal reduct based on distinction table. The algorithm can obtain the best solution with one chromosome in a short time. It is testified by two experiments that our algorithm improves the GA from four points of view: population size, parallelism, computing time and search capability.