New approaches to develop effective recommendations from large bank datasets are valuable and in the focus of commercial companies.
The purpose of this report is to document the attempt of the application of data mining methods for recommendations of bank products for specific customers.
The methodological approach is to explore whether product recommendation strategies can be achieved using three different methods, i.e. cluster approach, PCA and compare these with basket analysis results. The cluster approach and PCA can be combined for visualisation purpose. PCA also adds to the value as it can be additionally be used for dimensionality reduction.
My finding is that there is potential in this approach as it generates results that can be efficiently understood by domain knowledge owners, which is important for further development and refinement.
Conclusion is that using clustering combined with PCA/MCA opens doors for addressing specific recommendations to groups of clients. Further time should be invested into combining results of basket analysis results with those from clustering / PCA-MCA.
- R