PSO-Based Attribute Reduction of Rough Set and Its Application

摘要

Background: The rough set theory is a powerful tool to deal with imprecise and incompleteinformation in the field of data mining. As a core content for rough set theory, attribute reductionaims at removing redundant data and drawing the minimum attributes while maintaining indiscernibilityrelation. However, traditional rough set theory is available for classical example whichhas disadvantages of time-consumption, large storage and low recognition accuracy. In this paper,we focus on an attribute reduction based on particle swarm optimization (PSO) to overcome thedrawbacks of traditional rough set theory. Firstly, this paper reviews some important concepts ofrough set and particle swarm optimization. Then, we establish the model of attribute reduction basedon particle swarm optimization. Finally, the proposed method is applied to actual oil logging data,and the reduction results are recognized by Relevance Vector Machine (RVM) and Second OrderCone Programming-Relevance Vector Machine(SOCP-RVM). The experimental results show thatthe proposed method is efficient and has high recognition accuracy.Methods: Recent publications and patent databases are reviewed to find extraordinary and innovativeattribute reduction algorithms for reducing time consumption and accuracy.Results: Two methods which are RVM and SOCP-RVM are applied to recognize the attribute reductionresults. The results show that well in 993-997m, 1045-1152.5m and 1236-1255m depth aremain oil-layers, the rest are dry-layers (Fig. 2 shows that oil-layers are 995-997m, 1045-1152.5mand 1241.5-1255m; Fig. (3) shows that oil-layers are 993-996m, 1055.5-1143m and 1236-1251m).The recognition results are consistent with the actual oil test results.Conclusion: A novel PSO-Based Attribute Reduction of Rough Set is proposed, and apply it to oilwell to deal with actual and complex data. Experimental results show that the proposed algorithmcan get effective reduction sets, and the recognition results with high accuracy can be obtained inactual well by using RVM and SOCP-RVM algorithms, which are consistent with the actual oil conclusion.It indicates that the proposed attribute reduction based on PSO is practical and viable, andthe reduction results are efficient.