A developer's perspective of a decision support system

The article discusses the development of a data-centric decision support system (DSS) for process industry applications. DSS is defined as a computer-aided decision environment, positioned at higher levels of the control hierarchy, where the complexity of a task or the importance of a decision calls for human supervision. The data-centric DSS was developed when none of the traditional forecasting and optimization approaches were applicable to a municipal district heating system. The designated R&D team adopted a hybrid approach in which a multiple regression model was applied to a subset of past data points. The idea of data-centric forecasting involves viewing historical data records as points in the associated data hypercube. For any query point, defined by specific values of the predictor variables, a set of similar data points in the neighborhood are retrieved and fit with a statistical model. The model is used to calculate the forecast of the response variable and then the local model is discarded. The idea of data-centric optimization involves viewing the historical data as points in the associated data hypercube. Starting at the current state of the process, a set of actions is evaluated in a relevant neighborhood of the current state. New actions are searched for in the vicinity of the past best practices while attempting to improve the objective function. The paper continues with issues relating to the selection of a suitable technology, the transfer of knowledge from the development team to a business unit, and the practical aspects of DSS application that add to the cost of operation are discussed.