Hierarchical variable clustering based on the predictive strength between random vectors Int. Journal of Approximate Reasoning 170