Some Applications Of Matrix Completion Totext Classification

There is a lot of new data in the Internet, one of the most important formis text. We will do the text classification for data mining with theinformation.There will be correlation between two texts, for example, hyper link is aspecial form of Internet information. These correlations tend to revealinformation about the text category. In many conditions we can’t get all therelevant information between the texts, not only hyper link, but also otherforms. Always there is only correlations between the text and a few other texts(not necessarily in the same class!), and the degree of correlations can bedifferent. So we can use the text set to make a connection matrix, the elementat the crossing of the row and the column is denoted by the correlationbetween texts indexed by the row and the column. Thus the link matrix hasknown value in each row and column, but most of the component is unknown,then the unknown part of the matrix can be completed.Classifying by thecompleted matrix is the next work. We can complete the matrix effectivelybecause the classes which we will divide the texts into and the factors whichaffect the classifying are finite. The experiments show the matrix completionand cluster are available for the rows and columns if we get the information ofthem.