Evaluation metrics


As described previously, the problem of personality recognition consists in a binary classification problem for each of five traits. Thus, the evaluation of the participants teams will be through standar set-based metrics, such as F-measure, accuracy, and the the ROC AUC measure. The leading evaluation measure for ranking participants is  the macro averaged Area Under the ROC Curve (auc_multilabel)

**Important**  Submissions must be formatted in the same way as the sample submission (Files section-> Develpment button): each row in the submission file has predictions for each subject in the same order as they appear in the provided *.data file. Also, the prediction’s order for each row is important, that is, the first column will be consider the answer for EXT (extraversion trait), second column for AGR (Agreeableness trait), third column for CON (Conscientiousness trait), the forth for STA (emotional stability trait) and the fifth column for OPE (Openness trait). Notice that columns are separated by a single blank space. For the submission file, remember the file's name should start with HWxPIs_valid (for the validation partition) and must end with .prediction.


Challenge started!

The challange has started, please visiti the track's sites and the corresponding CodaLab pages for further information!

DivFusion track https://competitions.codalab.org/competitions/18419


HWxPI track https://competitions.codalab.org/competitions/18362