Implementation Of News Classification System Based On The Naive Bayesian

The explosive growth of information in the past few decades has generated ademand for technologies to collect and archive vast amount of news in an efficientmanner, because it would be practically impossible to do so in a manual manner dueto the concerns about manpower, money and time needed.This dissertation presents a brief survey on news classification techniques. Thepros and cons of Bayes classification algorithm has been discussed, before theincremental learning algorithm based on Na ve Bayes classification is studiedtogether with the CHI method for feature extraction. A new approach to choose textapproach based on incremental learning is then presented, based on which aprototype system is developed in Java. This system covers the whole range of newsclassification, such as text preprocessing, feature extraction, incremental learning,classifier and performance appraisal. Its performance has been tested with raw datafrom Nanfang Daily news archive.