Research On Optimal Scale Selection In Multi-scale Information System

In the era of "Internet plus",data is showing explosive growth at an extraordinary speed.How to get valuable information from it is a key issue in current knowledge discovery research.Multi-scale information system is a system which has many different attributes values due to the measurement of attributes in different levels of scale.It has become an important method to analyze and process data under the attribute information of different levels of scale.The decision results of different levels of scale may be different,so how to choose the optimal scale is a hot issue in the research of multi-scale decision system.However,there are still some problems worthy of further improvement in the existing optimal scale selection model.For example,when choosing the optimal scale,most existing models are from a single point of view,such as consistency and uncertainty,while the cost information widely exists in practical applications.In addition,the cost information of attributes is ignored in the optimal scale combination.Therefore,this thesis studies the above problems from the following two aspects:When all attributes have the same number of scales,that is,all attributes have the same importance,and the scale of attributes increase synchronously.On the basis of the sequential three-way decisions model based on the multi-scale decision system,the uncertainty is quantified.By introducing cost-sensitivity,a novel multi-scale decision system based on test cost and delay cost is defined,and then a cost structure based on the sequential three-way decisions model is proposed.Furthermore,the relevant change rules are discussed.By introducing the concept of ideal scale solution,an optimal scale selection model which combines uncertainty and cost-sensitivity is established.The experimental results show that the proposed model is more in line with the user requirements in practical applications.When attributes have different number of scales,that is,the attribute significance is different,and attributes can be combined with scales.Considering the influence of attribute cost information in practical application,the scale cost and attribute cost are defined by describing the cost of scales and attributes.By considering the two factors of attribute distinguishing ability and cost information,the weighted attribute significance is proposed to rank attributes.According to the attribute sorting sequence,select the optimal scale for each attribute until the optimal scale of all attributes are selected,and the result is the optimal scale combination.The experimental results show that the proposed model can select the optimal scale combination by considering the attribute distinguishing ability and attribute cost.To sum up,the cost-sensitivity is introduced into the multi-scale decision system to select the optimal scale to satisfy the user requirements in practical applications.