Study On Extended Rough Set Models And Application In Water Eutrophication Assessment

Rough set refers to a set of theoretical methods proposed by Polish Prof.Z.Pawlak for studying the expression,learning and induction of incomplete uncertain knowledge and data.Great progresses have been made both in theoretical research and practical application since it was proposed in 1982.Yet there are some limitations in the practical application of discretization,knowledge reduction as well as rule knowledge extraction,inference and decision making entailed in the classical Rough Set theory,such as:The low visualization and unstable partition of discretization intervals commonly exist as far as the traditional discretization algorithm is concerned;In the incomplete information system,it is difficult to satisfy the equivalence relation constraints of the traditional Rough Set,and the anti-noise capacity is weaker;The matching of traditional serial knowledge is relatively inefficient;In classical Rough Set,equivalence class is used to define an undefinable.static uncertainty set X through an upper and a lowerapproximation set,making it difficult to cope with the dynamic data mining and dynamic knowledge discovery in dynamic information systems.In this dissertation,the above four limitations are combined with the water eutrophication assessment in eco-environment fields to carry out and further the research on Rough Set model,the main innovations of which are as follows:1.A visualized discretization algorithm based on primary colors is proposed.A visualized discretization algorithm based on primary colors is proposed by referring to its principle in chromatology,in order to cope with problems that will occur when the traditional discretization algorithm is applied to the process of binary objects such as low visualization,missing information after discretization as well as complex calculation caused by excessive discretization.The results of contrast experiment conducted on the UCI public data set show that the proposed algorithm not only is able to realize stable and accurate data discretization,but also exhibits a good visualization and classification effect.2.A knowledge reduction algorithm for variable precision Rough Set in incomplete target information system based on variance relationship is proposed.The equivalence relation constraint of traditional Rough Set is hardly satisfied in incomplete information system,and the processing of incomplete information also requires a strong anti-noise capacity.In this dissertation,a variance relationship in incomplete target information system is put forward,based on which a variable precision Rough Set model is established.Corresponding definition and algorithm of knowledge reduction are then proposed,thus realizing the knowledge acquisition in the noised incomplete information system.The algorithm has now been applied to the knowledge reduction of the incomplete eutrophication data of the Xiangxi River in the Three Gorges Reservoir area,and the eutrophication knowledge acquisition in the noised incomplete information system has been realized.3.A parallel inference model that can be run in the incomplete information model is established.Aiming at the low parallel inferential capability of the traditional Rough Set model,a parallel inference model that can be run in the incomplete information system is established by referring to the extraordinary capability of Petri Nets.It has been applied to the knowledge inference of incomplete eutrophication data of Xiangxi River in the Three Gorges Reservoir,thereby realizing the efficient knowledge inference in the noised incomplete information system.4.Two algorithms based on two-direction S-Rough Set are put forward to acquire dynamic knowledge.Regarding the deficiency of using classical Rough Set to process dynamic knowledge,two algorithms based on two-direction S-Rough Set are proposed to acquire dynamic knowledge by referring to the element and property migration in two-direction S-Rough Set,in accordance with the particle size of element dynamic changes.The comparison between these two algorithms and other similar algorithms conducted on the UCI public data set indicates that these two algorithms have better classification accuracy and require shorter processing time.At last,guided by the dynamically-expanded Rough Set with two-direction multi-element migration,the precursory anomalies of the two algal blooms occurred in the spring of Xiangxi River located in the Three Gorges Reservoir area are analyzed,thus providing references for the forecast of algal blooms in Xiangxi River.In conclusion,the theoretical expansion of Rough Set model as well as a series of application research on the water eutrophication assessment on the basis of expanded Rough Set conducted in this dissertation not only facilitates the progresses in terms of knowledge discovery theories and approaches regarding Rough Set,but also provides a useful technical reference for the sustainable development of aquatic ecosystem and water quality management.