We proposed two metric approximations of the SProt measures. These approximations show (in combinations with the original SProt measure) good real-world results, which was tested on the ability to solve the classification problem.
Further experiment shows that the proposed measures should be suitable for metric indexing.