| Outlier detection is an important research content of data mining.It aims to distinguish objects with significantly different characteristics in the data set.It is applied to all aspects of production and life,and has theoretical research value and application prospects.However,the existing methods do not combine attribute sequences to detect outliers on the compatible rough set,and the outlier degree of data objects in data sets often involves the sequence relationship of multiple attributes and the impact of their integration.To solve this problem,this paper will make use of the advantage of compatible rough set model that can better deal with incomplete or uncertain data for knowledge discovery,and study the outlier detection of compatible rough set sequence.The main contents are as follows:(1)In order to investigate the influence of attribute sequence on the classification of data objects in the universe,the attribute sequence and an attribute set sequence are constructed by using the importance of attributes.According to the compatible relationship of compatible rough set,the compatible class of data objects in the dataset associated with the attribute subset is obtained,and then the compatible class sequence of objects is obtained by combining the attribute set sequence.The outlier degree of the data object is characterized by constructing the outlier factor of the compatible class sequence of the data object,and then the outlier detection method based on compatible class sequence(CCSOD)is studied.The comparative analysis of experimental results from the UCI dataset show that CCSOD is better detected in the group.(2)In the compatible rough set,the compatible entropy is a measure of uncertainty that describes compatible classes in information system.Based on the single attribute,the relative compatible entropy sequence outlier detection method(CESOD)is constructed on the attribute sequence.The experiment is conducted on a UCI dataset to test the effectiveness of the algorithm. |
