Research And Improvement Of ABLE-based Machine Learning System

Multi-agent system (MAS) is considered as a new and promising approach to putartificial intelligence technology into practice. In order to fully explore itsproblem-solving potential, learning ability must be properly integrated.The paper starts with a brief review of the basic concepts and features about agent,and that of MAS integration. Agent Building and Learning Environment (ABLE) is aJava-based framework for intelligent agent development and deployment. After a detailedanalysis of the machine learning and XML technology, this paper presents an XMLinterface which helps ABLE to read data in various formats. In addition, the decision treealgorithm used in ABLE is improved in order to classify and learn from data in an easierand more effective manner. An ABLE-based machine learning system based on MAS andmachine learning technology is then proposed. This system extends the basic classesprovided by ABLE and can learn from data effectively, with the help of the XMLinterface and the improved decision tree algorithm. The system is then tested using theUCI data sets and the experimental results analyzed, before a brief discussion aboutfuture work and some concluding remarks are made at the end of the paper.