H. Kou, G. Gardarin, A. D’heygère, and K. Zeitouni (France)
k-NN, document categorization, machine learning, XML
We have implemented the document categorization system DocCat to automatically organize French financial news for Firstinvest site. This paper describes system framework and main techniques we use. In DocCat, both relational database and XML are used to organize documents, our CBA algorithm is conducted to select features and k nearest neighbor algorithm is implemented as categorization model. We use 4000 financial news to learn and evaluate DocCat. The primary experimental results show that DocCat produces satisfactory performance. The flexible design allows users to easily adapt DocCat to different application domains.
Important Links:
Go Back
