Beginning Anomaly Detection Using

IRU

I

LQ

UDQJH [BYDOVKDSH>@ 

LI

SUHGV>I@   

VFRUH VFRUH

DFFXUDF\ VFRUH[BYDOVKDSH>@

SULQW

$FFXUDF\^`IRUPDW DFFXUDF\

Figure 2-52.  Generating the accuracy score for the validation set

Chapter 2   traditional Methods of anoMaly deteCtion

---

69

This time the accuracy was even better at around 89.5% (Figure 

2-53

).

Now to test on the novelties data set. This time, you can find the AUC score because 

there is a 50-50 split between anomalies and normal data. The other two data sets, x_test 

and x_validation, only had normal data, but this time it is possible for the model to 

classify false positives and true negatives.

Run the code in Figure 

2-54

.

Figure 2-53.  The resulting percentage of data points in the predictions that were 

considered normal