Research On The Estimation Of Null Values In Relational Database

Nowadays, the model of database system has been developed very vell and almost all aspects of DBMS have been very successful. The traditional database system is based on the complete information and does not allow the existence of incomplete information. But now the imcomplete data and null values are widespread in relational database. In general, a database system will not operate properly if containing null values of attributes and it is necessary to cope with them and that involves the problem of estimating null values.In order to solve those problems mentioned above, we propose three methods to estimate null values in relational database. First we use probability and statistics method and K-means clustering algorithm to improve the efficiency of algorithm. Second we propose a method which utilizes membership function to transform and recluster data. This method can not only estimate numeric values, but also estimate non-numeric values in relational database. Finally we use tuples which are compatible for each other to estimate null values. For verifying our methods, we use the mean of absolute error rate as an evaluation criterion to compare with other methods. It shows that our proposed methods are superior to some other methods for estimating null values in relational database systems.