Non-negative Matrix Factorization And Its Application To Fuzzy Classification Of Web Pages

Based on the statistical word frequency on the webpages and text classification methods rarely consider the fuzzy semantic word classification problem, so when the text in the presence of large semantically ambiguous words, classification of the effect is not ideal, so the paper introduces the fuzzy reasoning to solve the above problems In addition, this paper relates to dimension of the large scale term-document matrix, in this thesis, a large-scale text data is abstracted to a higher dimension matrix VSM, and deal directly with a higher dimension matrix data will be very cumbersome, imagine if large-scale text data is simplified as a low-dimensional matrices, the problem would be simple. Non negative matrix factorization algorithm is a kind of high order matrix dimensionality reduction method, has a simple, interpretable advantages, which can keep the NMF algorithm is applied to the matrix dimensionality reduction. Based on the text classification and NMF in the matrix dimensionality reduction on the advantages, this paper presents a non-negative matrix factorization based fuzzy webpage classification algorithm. The dimension of the large scale term-document matrix is reduced by NMF to condense the data and increase the execution efficiency and a fuzzy classification is used to design the classifier The results of classification experiments with the randomly selected webpage and the comparative experiments to singular value decomposition and General classification algorithm method show that the proposed algorithm has faster computing speed and better classification accuracy than that of others. Based on the above contents, the main work of this paper is as follows.Analysis of the current webpage classification generation, development process Summarize the current webpage classification of the main steps and methods. According to the text and webpages are common in both text, the text classification method is applied to the webpages classification, and Fuzzy inference rule is introduced to the webpage text classification process. The experimental results show that in the process of classification fuzzy inference will greatly improve the classification accuracy; and compare to the singular value decomposition (SVD) method, the implementation of efficiency has greatly improved.