lederb opened a new pull request #1496: URL: https://github.com/apache/systemds/pull/1496
This feature implements the Matthews Correlation Coefficient (MCC) calculation as a metric for binary classification evaluation. The selected builtin function name is "mcc" and its signature allows three modes of calculating the MCC: * one vector for predictions and one vector for labels, entries should be binary but are tested. If they are not binary, values are interpreted as lower value -> 0, higher value -> 1 * alternatively a 2x2 confusion matrix can be used containing TN, FP, FN, TP counts in row major order * option three is the calculation based on using named params TN, FP, FN, TP as parameters for the mcc call directly. Currently it contains a recursive call so checks on the confusion matrix are performed even if input were two vectors for predictions and labels, can be easily removed if unwanted. There are also some tests included, where some of those are expected to fail due to invalid input, the ones not expected to fail are compared against the result of a R mcc implementation. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
