Research On Ordered Decision Based Confidential Dominance Relation Rough Set Model And Its Application

Uncertainty of information is widespread in real world. Rough set theory is one useful tool for dealing with uncertainty information excellently. Compared with some other theories like probability theory, fuzzy set, etc., it is significant advantage that rough set theory needs not any prior knowledge and describes the problems objectively. In order to describe the preference order of attribute values, dominance-based rough sets approach was proposed to address multiple criteria decision problems. In real application, due to data missing, the study of ordered decision with the incomplete information is attracting the more and more attention. And it has been becoming an important field of intelligent information processing.Water quality evulation is one of the most important problems of water resource management and decision analysis in water ecological environment online monitoring system of Three Gorges Reservoir Region. The water quality data is uncertainty, incomplete and ordered. Hence, incomplete ordered decision analysis based on rough sets will provide theoretical support and the effective method for water quality assessment problem.In this dissertation, rough set model based on confidential dominance relation is proposed on process incomplete ordered information. Incremental updating approximations methods, sorting decision and attribute reduction in confidential dominance relation rough set are deeply studied and applied to water quality evaluation as new effective way. The main contributions of the dissertation are listed as follows.(1) The confidential dominance relation and its rough set approximation model are presented for decision analysis in incomplete ordered information, which fits the characteristics of the ordered relation to avoid semantic contradiction. Compared with the current models, confidential dominance relation based rough set model achieves the better approximate accuracy and better classification accuracy.Confidential dominance relation rough set model, which is suitable for water quality monitoring data, is proposed to incomplete ordering information processing. Various extended dominance relation rough set models are already proposed to solve the incomplete ordinal decision problem. But there exits some ambivalence over the real semantics because the characteristics of the order relation are not considered. Confidence dominance relation,which is more practical in the field of monitoring, fits the characteristics of the order relation to avoid semantic contradiction. Then uncertainty measure in incomplete ordering decision system is given. Furthermore, relationships and classification performance between the confidential dominance relation based rough set model and existing models are discussed. The experiment results illuminate that confidential dominance relation based rough approximation achieve the better approximate accuracy and better classification accuracy. The research achievements will be made for sort decision and knowledge in incomplete ordering decision system. (Chapter 2)(2) Methods of incremental updating approximation are proposed in confidential dominance relation rough set under variation of attribute sets or objects. The methods overcome the defect of recalculate the approximation in non-incremental updating and improve the performance of knowledge acquisition.Computation of approximations is a core issue in confidential dominance rough set model. In real-life applications, the attribute set or the object set is dynamically changed.Classic method of approximations calculation focuses on static data so that non-incremental updating is low efficiency for recalculating the whole space. Firstly, according to the variation of the attribute set, confidential dominance and dominated class are calculated.The principles of incremental updating approximations are discussed when some attributes are added or deleted, by which incremental updating approaches and algorithms are obtained under variation of attribute sets. Secondly, the principles of updating decision classes, confidential dominating and dominated classes are discussed when one object added or deleted and subsets of objects merged in incomplete ordered information system.The approaches to incremental updating approximations in confidential dominance relation rough set under the variation of the object set are proposed. The results demonstrate the proposed incremental approximations updating approaches are effective and efficient than non-incremental under the variation of the attribute sets or the object set. Research results can be applied to process dynamic monitoring data using confidential dominance rough set model (Chapter 3).(3) Fault-tolerance preference sorting decision models based on confidential dominance relation rough set are proposed to make the uncertainty ranking objects to achieve proper grade or nearest correct grade of the object, which are novel sorting methods to handle incomplete ordered decision system.The sorting decision problem is prevalent in daily life, such as water quality ranking,pollution level, etc. which has a decision attribute on preference ordered relationship different from classification problem. Fault-tolerance preference sorting decision models based on confidential dominance relation rough set are proposed by regarding the user's fault-tolerant preference such as upward, downward, or synthesis of the both. Two strategies are adopted to process the coverage information on the boundary: one strategy deems the coverage as the redundant or interferential information so that the ranked objects do not adjust their grade; the other strategy is to adjust the sorted results using the coverage information as the heuristic information. The experimental results of wine quality dataset show that the proposed methods to achieve better results on the correct rate and the mis-sorting cost that the model is effective in sorting decision. Research results can be applied to water resource management using confidential dominance rough set model(Chapter 4).(4) Attribute reduction approach based on inconsistent confidential dominance relation is proposed to obtain all reducts in incomplete ordered decision system. Furthermore,heuristic reduction and incremental reduction approaches are proposed to improve the efficiency.In order to discern two objects in ordered decision system, their decision preference order should be taken into account. Attribute reduction approach based on inconsistent confidential dominance relation is proposed, with which two objects are discernable. The judgment theorems and the discernable matrix are investigated, from which we can obtain a new approach to knowledge reduction in ordered decision system. An example illuminates effectiveness of the reduct. To enhance the efficiency of reduction, heuristic reduction algorithm is proposed by preserving quality of classification due to monotonicity.Furthermore, the principles of incremental updating approximations are discussed when one attribute is added or deleted. Incremental reduction algorithm is proposed by updating quality of classification. Finally, the experiments results show the two algorithms are effective, and incremental reduction algorithm is more efficient with bigger sample size.Attribute reduction is available to find out the important monitoring indexes (Chapter 5).(5) Confidential dominance rough set model is applied to water quality evaluation.Firstly, water quality and eutrophication level is graded by fault-tolerant sorting decision;then, the important monitoring indexes have been found by attribute reduction; finally,incremental updating approximation approaches improve the efficiency when the monitoring index or monitoring data dynamic changes.Research results above-mentioned have been fully applied to water quality evaluation.A case study of Fengshouba water monitoring data, fault-tolerant sorting decision approaches are used to determine water quality grade and eutrophication level. The experiment results show the approaches are effective. Furthermore, the important monitoring indexes can be analyzed by attribute significance and the more concise indexes set can be obtained by attribute reduction. To enhance the efficiency with monitoring data or indexes dynamic increasing, incremental updating approximation methods are applied to water quality evaluation. The experiment results on Fengshouba water monitoring data show ordered decision methods based on confidential dominance relation rough set model are effective. And the approaches are scalable when monitoring indexes or evaluation benchmarks are changed (Chapter 6).