Dominance Lagrange Optimized Rule Generation for Decision Table Evaluation

Manipulating the volumes of data and recognizing significant information was the challenging task in organizing and managing with uncertain data. In data management, evidence theory and fuzzy measure evaluated the uncertain information for decision execution from an arranged decision table. Despite, improvement in the decision performance and validity, an optimized rule generation has to be studied. Through this paper, a Dominance Principle-based Reduct with Dominance Lagrange Optimized Rule Generation (DLO-RG) model is proposed for the analysis of obtained decision rule set. In DLO-RG, the degree of coherence to the dominance factor is defined. Then, the coherence degrees of all objects are combined, called combined valued preference relation dominance factor to generate the reducts of an input decision table. To the generated reducts, a Deterministic Global Optimization model using Lagrange is designed to optimize the rule generation. Tight lower bounds and upper bounds are generated using Lagrange formulation. The Deterministic Global Optimization model contributes optimum lower bounds, eliminating uncertainty while extracting rules and therefore improving rule extraction accuracy. Theoretical analysis conducted using the DLO-RG proves that the extracted reducts have lower redundancy, therefore reducing the rule generation time and are more diverse when compared with those obtained by existing evidence theory. Experimental results prove the rules framed by the DLO-RG are more significant when compared with that of rules from existing methods.