New M.E.Thesis Submitted from IT

ONLINE NEWS CLASSIFICATION USING NEURAL NETWORK AND SUPPORT VECTOR MACHINE CLASSIFICATION By Bhumika. 

Abstract

News classification is the growing interest in the field of text mining. Availability of different types of dataset made the online news classification a crucial topic in today’s era. Support Vector Machines (SVM) and Neural network (NN) are good approaches in the context of classification problem. Text organizing has become very important in an organization too to maintain the different types of similar data.  Therefore, there is a need of automatic text classification system. In proposed work use of SVM and NN for on-line news classification has been done with the aim of better precision and recall rate. Within the proposed work, three forms of news has been classified like financial, sports and political. In this thesis, the main objective of text classification is the classification of documents into a settled amount of predefined classes. Every single archive could be in no class, several, or precisely one by any means. Our main goal is to utilize machine learning algorithm to understand classifiers. From cases which execute several categories assignments subsequently. This is type of issue is a supervised learning issue. Since classes might possibly cover, every specific class is dealt with as a different double order. The initial phase in content characterization is to alter records, which generally are series of characters, into a representation which is appropriate for the learning algorithm along with the arrangement assignment. Data Retrieval research anticipates that word stems function quite well as demonstration units and that their sequence in a text document file is of small importance for several tasks. This directs to an attribute value representation of text.  The proposed work implementation includes the use of SVM and NN for online news classification. In the proposed work, three types of news has been classified like financial, sports and political and the whole implementation has been taken place in MATLAB 2010a.From results simulation it has been seen that NN accuracy comes out to be better than SVM. Also in the end the proposed work has been compared with nave bay’s classifier and it has been concluded that the accuracy of proposed work is better having value 99% then previous technique having value 92%.