Decision Making Using Fuzzy C-means and Inductive Machine Learning for Managing Bank Branches Performance

Decision making tools and architectures of nowadays consist an effective way of extracting useful information from complex data domains. A decision making scheme is presented here, combining fuzzy cluster analysis and TDIDT (Top Down Induction of Decision Trees) algorithms. Cluster analysis is primarily used by applying fuzzy C-means, in order to group uniform data sets stored in databases as attribute value vectors. This process has two outputs: an identification of clusters (through description of clusters’ “center”) and a membership function to each attribute value vector. Once clusters have been identified, attribute value vectors are extended, containing the membership value of the data set to each cluster. An inductive learning algorithm is then applied, producing meaningful discrimination rules among clusters, in the form of a decision tree. The proposed decision making architecture can cope with any kind of data quantitative or qualitative depending on the application domain and the nature of the decisions to be taken. A real world application has been selected to demonstrate the effectiveness of our methodology. Specifically, it is shown how the top-management team of a domestic bank network should decide, about granting the annual productivity bonus among the branches of its network. Further directions are addressed through this paper, proposing the construction of a data mining tool, able to discover interesting information in on-line mode.