Execution Engine of Meta-learning System for KDD in Multi-agent Environment

摘要

Meta-learning system for KDD is an open and evolving platform for efficient testing and intelligent recommendation of data mining process. Meta-learning is adopted to automate the selection and arrangement of algorithms in the mining process of a given application. Execution engine is the kernel of the system to provide mining strategies and services. An extensible architecture is presented for this engine based on mature multi-agent environment, which connects different computing hosts to support intensive computing and complex process control distributedly. Reuse of existing KDD algorithms is achieved by encapsulating them into agents. We also define a data mining workflow as the input of our engine and detail the coordination process of various agents to process it. To take full advantage of the distributed computing resources, an execution tree and a load balance model are designed too.