====================================================================================
"Anomaly Detection for a Water Treatment System Using Unsupervised Machine Learning"
Inoue et. al, DMCIS 2017
====================================================================================

ERRATA (Updated 15th May 2021)

The dataset we used for training contained a small number of typos: 37 out of 449919
entries had their label misspelled as "A ttack" with a spurious space. All other la-
els were spelled correctly.

Unfortunately, our Python scripts interpreted these misspelled labels as "Normal" or
"Neither Normal nor Attack", depending on the line of code. This makes a small diff-
erence in the overall performance of the models. In particular, for 1-class SVM, our
paper presents the following performance:

      precision: 0.9249920360704746
      recall:    0.6990111477352691
      F-score:   0.7962789127843817
	  
whereas the current dataset on the iTrust website gives the following (with credit
to Sophia White, Kings College London):

      precision: 0.9836
      recall:    0.6737
      F-score:   0.7997