Document categorization in multi-agent environment with enhanced machine learning classifier

Text categorization task have gained the attention of researchers in last 10 years with the increase in web-based contents of documents. For searching a particular document from the web or any large document collection text or document categorization is most useful task. We demand some better system and enhanced machine learning classifiers to accomplish task of document categorization. We designed a multi-agent based system that consists of some software hybrid agents that obtains the category of a document and interact with each other to take final decision about the category and then data is fed to a machine learning classifier in order to enhance the performance. We analyzed the results of the system in form of performance measures such as accuracy, recall, precision and true negative rate. We analyzed the result in two scenarios: one is with decision of agents alone and another is with application of reinforcement clustering technique neural network. We observed that in first scenario the system's accuracy, precision and true negative rate is very good and recall measure is significantly good. After the application of reinforcement clustering technique there is no significant change in system's performance instead the recall is degraded. But still the system is producing good results in both scenarios.