A workflow-based framework was developed facilitating seamless use of available DataMiningGrid and other WSRF-compliant services and Web services. It is based on Triana. The client components can be used in a variety of ways, they are very flexible and allow for building and execution of novel, distributed data mining analysis and applications.

The workflow editor (see figure below) acts as the major end-user interface. It provides graphical creation, loading, and storing of workflows to be executed in the grid. As managing the various aspects of grids such as job submission, job monitoring, and error handling are inherently complex, the editor will provide significant support for the user by insulating him from the details these aspects. It will also help enforcing complex data mining tasks, which might consist of numerous different steps and usually include accessing, transformation, and pre-processing data as well as applying various Data Mining algorithms, evaluating the results, and storing the models on an appropriate server in the grid or on the client machine. The workflow editor is the only system component, which is installed locally on each client machine.

The above picture is an example of the workflow editor use. It represents the implementation of a data mining task (Weka’s algorithm J48) in a distributed grid environment. Here, the system is designed to solve relatively simple, but common task of finding parameters that enable the classifier to produce a model with an error rate below a certain threshold.

Updated on June 4, 2008