Beginning Anomaly Detection Using

Figure 7-39.  Code to start the training process for the model

Figure 7-40.  The output during the training process

Figure 7-41.  The output when the training process ends

Chapter 7   temporal Convolutional networks

---

281

The output should look somewhat like Figure 

7-43

.

Now you can check the AUC score (see Figure 

7-44

).

The output should look somewhat like Figure 

7-45a

.

VFRUH 7&1HYDOXDWH [BWHVW\BWHVW

YHUERVH 

SULQW 7HVWORVV VFRUH>@

SULQW 7HVW PDH VFRUH>@

SULQW 7HVWDFFXUDF\ VFRUH>@

Figure 7-42.  Code to evaluate the loss and the accuracy on the test sets

Figure 7-43.  The generated loss and accuracy scores for the test set. The accuracy 

is really good, but again, accuracy isn’t always the best metric to judge models by

IURPVNOHDUQPHWULFVLPSRUWURFBDXFBVFRUH

SUHGV  7&1SUHGLFW [BWHVW

\BSUHG QSURXQG SUHGV

DXF URFBDXFBVFRUH \BSUHG\BWHVW

SULQW $8&^`IRUPDW DXF

Figure 7-44.  Code to generate an AUC score given the test sets and the 

predictions

Chapter 7   temporal Convolutional networks

---

282

For the classification report and confusion matrix, see Figure 

7-45b

.

Figure 7-45a.  The generated AUC score of 99.02% for this model

Figure 7-45b.  Classification report and confusion matrix

Chapter 7   temporal Convolutional networks

---

283

That’s a pretty good AUC score! However, this was an example of