From: Predicting abnormal trading behavior from internet rumor propagation: a machine learning approach
Metric | Definition and Computation |
---|---|
Accuracy | The proportion of correctly classified records \(\frac{Number of correctly classified records}{{Total Number of records}}\) |
Sensitivity | The True Positive Rate (TPR); the proportion of correctly classified positive records (i.e., abnormal trading) among all positive records \(\frac{Number of correctly classified positive records}{{Total Number of positive records}}\) |
Specificity | The True Negative Rate (TNR); the proportion of correctly classified negative records (i.e., normal trading) among all negative records \(\frac{Number of correctly classified negative records}{{Total Number of records as negative}}\) |
Precision | The positive predictive value; the proportion of records correctly classified as positive (i.e., abnormal trading) among all records that were classified as positive \(\frac{Number of correctly classified positive records }{{Total Number of records classified as positive}}\) |
AUC ROC | Area under the Receiver Operator Characteristics (ROC) curve |