Designing a Knowledge Modeler Framework

Knowledge discovery is an iterative and interactive process where the appropriate solution for a given problem can be found only with big effort. There are several algorithms for each piece of problems that can be used for discovering hidden relationships in data. Choosing the best feasible method is not a trivial process. It is advisable to apply more than one method to the selected partition of the whole dataset in order to make a good decision. For this reason we developed a novel framework called Knowledge Modeler and Data Miner (KDMD) that can be used for testing several methods for a given problem in parallel. In this paper we show how the framework is designed for this purpose. The framework contains a running engine and model building environment that makes possible to build a complete knowledge discovery system by connecting small parts of processing blocks. The building blocks of the process are controlled by a data flow. The advantage of the framework is its easy extensible behavior that means adding a new component to the process is only implementing the algorithm itself using a well-known predefined interface to the input and output of the algorithm. Additional advantage of the framework is that it provides a common execution environment for the different algorithms thus comparing them can be done easily.