Skip to main content

Table 3 Comparing performance between conventional anomaly detection (CAD) and the proposed clustering approach

From: A clustering approach for detecting implausible observation values in electronic health records data

  1. * best performances are highlighted
  2. ** ties are in Bold
  3. *** best sensitivity among CAD methods was obtained from applying Mahalanobis Distances and 3.717526 (sqrt of 13.82) as critical value
  4. **** best specificity among CAD methods was obtained from using 6 standard deviations as threshold for identifying outliers