Neural Network Framework Components

The goal of this paper is to describe the design and implementation aspects of a framework architecture for decision support systems that rely on artificial neural network technology. Besides keeping the design open for supporting various neural network models, a smooth integration of neural network technology into a decision support system forms another important design goal. Many conventional implementations of such decision support systems suffer from a lack of flexibility, that is, they are built for a particular application domain and rely on one specific algorithm for the intelligent engine. In general, for different application domains, large portions of the decision support system have to be reimplemented from scratch. The principal contributions of this paper are: the description of flexible and reusable components for core aspects of neural networks implementations, the integration of different neural network models in a decision support system, and the presentation of a decision support system architecture that can be easily adapted to handle different domain problems. The chapter first outlines the flexibility problems of a typical conventional implementation, and then goes on to discuss in detail the overall architecture of the object-oriented redesign together with some relevant implementation aspects.