Ability to collect and utilize historical data is very important for efficient decision support. On the other hand this data often requires significant pre-processing and analysis in order to bring any value for the user-decision-maker. Knowledge Discovery in Databases (KDD) can be used for these purposes in exploiting massive data sets. This paper describes a computational architecture for decision support system, which comprises an artificial neural network component for the KDD purposes. It integrates mining data set stored in databases, knowledge base produced by data mining and artificial neural network components, which serve the role of an intelligent interface for producing recommendations for decision-maker. The architecture is being implemented in the context of aviation weather forecasting. The proposed architecture can serve as a model for a KDD-based intelligent decision support for any complex decision situations where large volume of historical data is available.