Beginning Anomaly Detection Using

 Anomaly Detection with the Dilated TCN

Now that you know more about what a TCN is and how it works, let’s try applying a 

dilated TCN to the credit card dataset.

First, import all of the necessary packages (see Figure 

7-18a

).

Input Layer

Hidden Layer 1

Hidden Layer 2

Output Layer

Figure 7-17.  A causal dilated network. The first hidden layer has a dilation factor 

of two, and the second hidden layer has a dilation factor of four. Notice how no 

inputs forward in the sequence contribute to the next layer’s node. This type of 

structure is ideal if the goal is to preserve some sort of flow within the data set, 

which is time in our case

Chapter 7   temporal Convolutional networks

---

268

Then, you must create a class for the visualization of confusion matrix, etc.  

(see Figure 

7-18b

).

Figure 7-18a.  Importing all of the necessary packages in order to start your code

Chapter 7   temporal Convolutional networks

---

269

After that, proceed to importing your data set and processing it (see Figure 

7-19

).

Figure 7-18b.  Creating a visualization class

df = pd.read_csv("datasets/creditcardfraud/creditcard.csv", 

sep=",", index_col=None)

print

(df.shape)

df.head()

Figure 7-19.  Importing your data set and displaying the first five entries

Chapter 7   temporal Convolutional networks

---

270

The output should look somewhat like Figure 

7-20

.

The data frame continues in Figure 

7-21

.

Each entry is noticeably large, with 31 columns per entry. If you check the tail end of 

the data frame in Figure 

7-22

,