Research On Attribute Reduction Algorithms Based On Rough Set Under Attribute Group Order

The development of social informatization makes the data increase rapidly.Extracting useful information from the huge data is the key problem of data mining,and the value of data is reflected in the different knowledge that different users need for the same data set.However,the common data processing methods only consider the information contained in the data set itself,which ignoring the external factors such as user needs.Therefore,how to mine the required knowledge from the data set for specific users,and how to adjust and optimize it combined with user preferences is a hot issue in current data mining.Attribute reduction is an important step in data preprocessing,which provides support for subsequent data processing methods.The purpose is to keep the classification ability of the original system roughly unchanged,and remove redundant attributes to achieve the effect of data dimensionality reduction.The algorithms of the existing attribute reduction based on rough set theory only consider the internal information of the data set,and ignore the different preferences of different users,so the established reduction model is not universal.Therefore,it is necessary to incorporate attributes into the algorithm design according to different user preferences,and establish a more practical and suitable reduction model for different users.As one of the user preference methods,the attribute group order is less difficult to obtain and is more widely used than attribute total order and balanced order.This article mainly studies the reduction problem under the attribute group order,which the content is as follows:1.The attribute reduction combined with cost-sensitive under the attribute group order is studied.Firstly,the attribute group order is used to represent user preferences.Secondly,based on the characteristics of attribute group order,attributes can be added or deleted by group.Then,the problem of attribute cost in reality is considered.For the attributes in the same group,the attribute importance and cost importance of attributes are combined to select by weighted method.Finally,the effectiveness and feasibility of the cost-sensitive algorithm under the attribute group order are verified,and considering the attribute group order and internal attribute information of information system,the attribute reduction can meet the needs of specific users.2.The dynamic attribute reduction for multi-user is studied.In order to meet the different needs of different users for reduction,a multi-user oriented dynamic attribute reduction method based on three-way decisions is proposed to adapt the change of user preferences.First,multi-user preferences are integrated,and the user preference matrix is defined to describe the preference of each attribute under multiple users.Then,combined with the attribute preference degree and the cost of practical problems,the user preference index is proposed,which represents the importance of attributes in the current user group and serves as heuristic information to select attributes.Finally,the attribute of the reduct set and the non-reduct set is divided into three parts by using the three-way decisions theory to achieve the purpose of updating the reduction.At the same time,the initialization algorithm of attribute reduction and the updating algorithm of dynamic reduction are proposed,and the reduction from the algorithms can meet the needs of multi-user.