Beginning Anomaly Detection Using

Figure 2-14.  The dimensionality of the filtered df

Figure 2-15.  The unique labels in df along with the number of instances of data 

points in df with that specific label

Chapter 2   traditional Methods of anoMaly deteCtion

---

43

You can also run print(df.head(5)), but it prints in a text format (Figure 

2-17

).

To resolve this issue, the 

label encoder takes the unique (meaning one entry per 

categorical value instead of multiple) list of categorical values and assigns a number 

representing each of them. If you had an array like

[ "John", "Bob", "Robert"],

the label encoder would create a numerical representation like

[0, 1, 2], where 0 represents "John", 1 represents "Bob", and 2 represents 

"Robert."

Figure 2-16.  A line of code to display the top five entries in the table. In this case, 

the image has been cropped to show the first few columns

Figure 2-17.  The same function as in Figure 

2-16

, but in text format

Chapter 2   traditional Methods of anoMaly deteCtion

---

44

Now do the same with the labels in your data frame.

Run the code in Figure 

2-18

.

encoded.fit(df[col]) gives the label encoder all of the data in the column from 

which it extracts the unique categorical values from. When you run

df[col] = encoded.transform(df[col])

you are assigning the encoded representation of each categorical value to df[col].

Let’s check the data frame now (Figure 

2-19

).

Good, all the categorical values have been replaced with numerical equivalents.

Now run the code in Figure 

2-20

.