Research On Filling And Attribute Reduction Algorithm Of Incomplete Inconsistent Data

With the rapid development of the Internet,the data shows the trend of polymorphism,which produces many incomplete and inconsistent data.Therefore,the processing of data classification and prediction has brought great obstacles.Filling the incomplete inconsistent data has always been a difficult problem in data preprocessing research in the field of data mining.Properly filling data can provide more useful information for data classification and processing.The process of high-dimensional data processing is also relatively complex.Effectively using attribute reduction can reduce high-dimensional data to low-dimensional data,which can greatly improve the speed of processing data and reduce the difficulty of analyzing incomplete and inconsistent data.At present,the rough set theory and related algorithms can well solve the problem of data filling.However,there are still many problems in the study of the filling algorithm and attribute reduction algorithm for incomplete inconsistent data.Therefore,this paper mainly studies the incomplete inconsistent data by combining information gain and inconsistent degree fill algorithm,and proposes an attribute reduction algorithm based on granular model,then improves the attribute reduction algorithm based on the granular model.The main work in this article is as follows:(1)Aiming at the computational time-consuming problem of calculating the equivalence class and the tolerance class in the attribute reduction,the tolerance relationship and the properties of the upper and lower approximations are used to build a granular model for quickly accessing and calculating the equivalence classes and tolerance classes of each object.The model can effectively shorten the attribute reduction time,and propose an attribute reduction algorithm based on the granulation model.After the granular model is built for incomplete inconsistent data,the information gain value of each attribute is calculated,and is sorted in ascending order to form a set of attributes to be reduced.Using on the consistent degree of attribute as the heuristic function of the attribute reduction algorithm,it computes and compares each object's degree of consistency with the attribute,then performs the attribute reduction operation,and verifies that the attribute reduction algorithm has better performance.Finally,the relationship between the ratio of inconsistency and the number of attribute reductions is also studied.(2)By combining the characteristics of information gain and inconsistency,this paper proposes a filling algorithm for incomplete inconsistent data.It can restore the characteristics of the original data to the maximum extent,and can achieve better filling effect and maintain a high classification accuracy.Then,an attribute reduction algorithm combining information gain and inconsistency is proposed for incomplete inconsistent data.The experimental results show that the attribute reduction algorithm proposed in this paper has better attribute reduction effect and expansibility,the attribute reduction is not only for the complete consistent data,but also for incomplete inconsistent data.In summary,this paper proposes a granulation model for incomplete inconsistent data and implements an attribute reduction algorithm that can effectively shorten the attribute reduction time,and then combines the information gain and inconsistency proposed by incomplete inconsistent data.Filling in the algorithm and the attribute reduction algorithm can restore the characteristics of the initial data to the utmost,and carry out effective attribute reduction.The experimental results prove the effectiveness of the algorithm.