Silhouette-based method for portfolio selection

Many new methods have been proposed for improving protfolio selection process, applying clustering techniques to stock price time series. We apply a recently proposed clustering method that employs stock price’s time series P-Spline coefficients, instead of the original time series, in order to speed up and stabilize the clustering process. Then we propose a cluster validation benchmark technique that enable us to automatically build financial portfolio that complies with some strategic constraints.