Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract： Attribute reduction is of great importance for data processing. For an attribute reduction algorithm based on the neighborhood rough set model, the calculation of the positive region is the necessary basis of its efficient performance and the uppermost part of its time cost. In order to reduce the time overhead of the algorithm, this paper improved the positive domain calculation of the existing algorithm FHARA, adopted the reservation strategy and used the matrix to preserved the square of the calculated values. The original n-dimensional computation was improved to 1 dimensional computation, which reduced the computation time of each metric calculation. On this basis, this paper proposed a neighborhood rough set attribute reduction algorithm based on the matrix reservation strategy. Finally, the algorithm was verified by multiple UCI data sets. Compared with existing algorithm, the experimental results show that for most data sets, the algorithm can get the attribute reduction of the dataset more effectively and quickly.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： The application of neighborhood rough set depends on the value of neighborhood size #1;. When using attribute reduction algorithms based on neighborhood rough set, an existing method for determining #1; is usually point-type, that is, to specify a value only by human experience. The method does not combine with the actual situation when it is used to determine #1;, so the practicability of the algorithms can be further discussed. For this reason, an adaptable method for determining #1; is proposed. The biggest characteristic is not determining #1; but the interval of #1;, then the most appropriate #1; in the interval is forwardly selected by using a fitness function that is combined with the characteristics of data sets and classifiers. The experimental results show that, compared with the point -type method for determining #1;, this method can find reduction sets which number of attributes is less, and classification accuracy is higher, which proves that this method can further improve the practicability of attribute reduction algorithms based on neighborhood rough set.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： Knowledge reduction of Pawlak rough set includes two parts: knowledge reduction for decision tables and knowledge reduction for information tables. As an extension of Pawlak rough set, neighborhood rough set is widely applied to attribute reduction for decision tables, but rarely applied to attribute reduction for information tables. In order to design an attribute reduction algorithm suitable for information tables, this paper first proposes a knowledge reduction criterion of neighborhood rough set for information tables based on the knowledge reduction criterion of Pawlak rough set. Then, according to this criterion, a new attribute reduction algorithm for information tables, applicable to clustering, is proposed with Greedy Strategy. Compared with Principal Component Analysis(PCA) algorithm, the experimental results show that by using this proposed algorithm to reduce dimensions of data sets, the number of attributes in the reduction sets is more, and the accuracy of K-means algorithm is higher according to the reduction sets, which proves this proposed algorithm can be effectively applied to attribute reduction for information tables.
Hits
Hits
Downloads
Comment