Beginning Anomaly Detection Using

Figure 4-19.  Code to evaluate the model

Figure 4-20.  Code to measure anomalies based on a threshold

Let’s delve deeper into the code shown above because this will be seen throughout 

the chapter when you classify data points as anomalies or normal. As you can see, this 

is based on a special parameter called the threshold. You are simply looking at the error 

(difference between actual and predicted) and comparing it to the threshold. First, 

calculate the precision and recall for threshold = 10. Figure 

4-21a

 shows the code to show 

the precision and recall.

Figure 4-21a.  Code to show the precision and recall

Chapter 4   autoenCoders

---

138

Let’s also calculate for thresholds = 1, 5, 15. See Figures 

4-21b

, 

4-21c

, and 

4-21d

.

Threshold = 1.0