Research And Application Of Stepwise Optimal Cut Selection And Dynamic Update Method Under Multi-scale Information System

Data processing under a multi-scale framework can satisfy the problem analysis from different perspectives,and the multi-scale rough set model promoted the development of multi-scale data analysis.Under this model,the selection of the most appropriate scale level for data analysis with different application needs has become an important research topic.Compared with the scale combinations under traditional models.,the cut based on granularity tree in multi-scale decision systems can enable different objects to take different scales under the same attribute,which realizes crossgranular knowledge acquisition.The semantic expression of cut combination based on the granular tree is more accurate.Generally,finer cut combinations will result in higher costs of knowledge acquisition,while coarser cut combinations can make decision rules more generalizable but may not fully satisfy decision-making needs.Therefore,how to select the most appropriate cut combination has become one of the research hotspots in the multi-scale rough set model.The cross-scale cut combination often leads to a large search space for selecting the optimal cut,which results in high time complexity of the algorithm.In practical applications,one optimal cut combination for data analysis and rule acquisition is usually demand.In addition,data in multi-scale decision systems under networked environments is often dynamic changing,and static methods can be inefficient in handling dynamic data due to duplicate calculations.To address these problem and requirements,this dissertation conducts research on the stepwise selection and dynamic update method of the optimal cut in multi-scale information systems,and explores the application of the optimal cut combination in practical production.The relevant research has not only improved the efficiency of selecting the optimal cut but also extended the research on the dynamic update of the optimal cut and promoted its application.The main research work of the dissertation includes:(1)The stepwise optimal cut selection algorithm based on multi-granularity attribute significanceThe cut based on granularity tree allows different objects to take different scales under the same attribute,and its cut combination is more accurate than the scale combination under traditional rough set models.However,due to the larger number of cut combinations,the search space for optimal cut selection is too large and the time complexity is high.To address this issue,this dissertation proposes a stepwise optimal cut selection algorithm based on significance order.Firstly,node significance,cut significance,and multi-granularity attributes significance under multi-scale decision systems based on granularity trees are defined respectively,and their internal relationships are studied.Then,the definitions of the optimal cut and the optimal node and their internal relations are given,and it is the first time that the optimal cut selection problem is transformed into the optimal node selection problem.Moreover,two stepwise optimal cut selection algorithms are proposed,which greatly reduces the time complexity of optimal cut selection.Finally,the algorithm is applied to the fresh ecommerce case,and the advantages of the decision rules extracted based on the optimal cut combination in generalization ability are discussed.Multiple sets of comparative experiments under the UCI standard datasets verified the effectiveness of the algorithm and its good adaptability under complex datasets.(2)Stepwise updating of optimal cut in multi-scale decision systems considering the dynamic changes of attributesIn practical applications under the networked environments,the data in information systems often changes dynamically.In order to solve the dynamic updating of the optimal cut in multi-scale decision systems,the stepwise updating method of optimal cut while attributes dynamic changing in multi-scale decision systems are proposed based on the static stepwise optimal cut selection algorithm.Firstly,an initial cut combination and attribute sequence are determined on the basis of an optimal cut combination and attribute sequence existing under the original systems,which avoids the recalculating of attribute sequence when updating the optimal cut dynamically.Secondly,two simplified theorems for node consistence judgement and a stepwise lower-bound cut selection algorithm are proposed,which can shorten the time of determining node consistence.Finally,stepwise dynamic updating algorithm of optimal cut with increasing and decreasing attributes are proposed respectively.Multiple comparative experiments on the UCI standard dataset show that,compared with the static stepwise optimal cut selection algorithm,the proposed dynamic updating algorithm can correctly obtain an optimal cut while significantly improving the computational efficiency.(3)Stepwise updating of optimal cut in multi-scale decision systems considering the dynamic changes of objectsConsidering the case of object dynamic changes in multi-scale information systems,this dissertation further studies the stepwise updating method of optimal cut while objects dynamic changing.Firstly,the impact of equivalent class merging on positive regions is discussed in multi-scale decision systems,which explains the impact of merging of attribute values that causes to equivalent classes to merge in the optimal cut selection process.Then,various situations that whether and how to update the optimal cut combination when only one object changes are considered,which leads to the general situation of multiple objects changes and relevant proofs are given.Finally,stepwise dynamic update algorithms for optimal cut considering objects added or removed are designed and applied to the chemical production to provide the guidance and suggestions for optimizing production and reducing costs for enterprises.