ABSTRACT
To address various requircments of data analysis in human performance and effectiveness such as Air Force Biodynamic Database, Navy Biodynamic Database, CAESAR database, etc., we developed a suite of flexible and scalable software tools, namely ABMincr (Agent Based data Miner). The goal of the program is to exploit large amount of bioscience sensor data, build abstract and general knowledge models from the data, and utilize those knowledge models for effcctivc decision making. ABMiner consists of three key components. First, ABMiner provides a flexible data screening component that extracts different datasets using the similar language of domain experts, not requiring the knowledge of database and SQL. Second, ABMiner provides an optimization engine in the meta-learning level to exploit and search data mining models with best performancc and efficiency among a wide range of data mining algorithms and their corresponding
parameters. To speed up the optimization searching process, ABMiner utilize IAI's well-established product CybelePro for distributed computation. Finally, ABMiner provides a model base component that enables a well-trained data mining model together with its underlying data set and performance metrics to be wrapped up and accessible across various platforms.
