This article discusses a design methodology for a Decision Support System (DSS) in the area of Data-Driven Management (DDM). We partition the DSS into an offline and an online system. Through rigorous testing, the offline system finds the best combination of Data Mining (DM) and Artificial Intelligence (AI) algorithms. Only the best algorithms are used in the online system to extract information from data and to make sense of this information by providing an objective second opinion on a decision result. To support the proposed design methodology, we construct a DSS that uses DM methods for market segmentation and AI methods for product positioning. As part of the offline system construction, we evaluate four intrinsic dimension estimation, three dimension reduction and four clustering algorithms. The performance is evaluated with statistical methods, silhouette mean and 10-fold stratified cross validated classification accuracy. We find that every DSS problem requires us to search a suitable algorithm structure, because different algorithms, for the same task, have different merits and shortcomings and it is impossible to know a priory which combination of algorithms gives the best results. Therefore, to select the best algorithms is empirical science where the possible combinations are tested. With this study, we deliver a blueprint on how to construct a DSS for product positioning. The proposed design methodology can be easily adopted to serve in a wide range of DDM problems.

This content is only available via PDF.
You do not currently have access to this content.