Skip to main content

Advertisement

Table 1 Classification results

From: Pool detection from smart metering data with convolutional neural networks

  Classification Method Accuracy Precision
a) All-positive 10.5% 10.5%
  All-negative 89.5% -
b) SVM Gaussian 93.1% 66.7%
  5-NN 94.0% 68.5%
  1-NN 93.4% 66.7%
c) CNN + 7-NN 93.1% 60.0%
  CNN + 5-NN 93.1% 57.7%
  CNN + 1-NN 93.4% 60.6%
d) CNN pure 95.5% 71.9%
  1. a) Baseline classification by classifying all instances as “pool” or “no pool”, respectively. b) Classification with manually designed features as in (Burkhart et al. 2018). c) Nearest-neighbor classification with CNN features. d) CNN classification
  2. Best performing algorithm in boldface