Page 108 - Code Craft Computer-8

of queries that help us achieve various objectives and better decision-making. These questions
can be divided into the following categories:

1. What would be the class of the given item?
In this category, a question related to the class of an item is asked, for example, what would be the

class of a given item, either A or B. For example , for the question "Will you go to the party
today?" the answer would be a yes or no.
In this category, the answer would be in some definite form. Some more examples of this type of
question are:

• Whether the tea is cold or hot?
• Whether it is a dog or cat?

• Whether it is an apple or a mango?
• Are you diabetic?

To give answers to such types of questions, classification algorithms are used.
Classification is a method which is used to classify the output into a finite number of classes. The
number of output classes may be two or more than two.
If the number of output classes is two, then the classification is called binary classification, and in

the case where you have more than two classes, it is called multi-class classification.

2. Is this an outlier?
An outlier means an object whose features are not similar to those of
other objects in a particular group. Suppose you have a basket containing
a lot of raw and green mangoes. If you find a ripe and yellow mango in
the bag, then that would be an outlier, because its features will be
different from the rest of the mangoes.

The concept of an outlier is used in an anomaly detection. A basket
An anomaly means something that is not normal or is different from the other objects in the same
group.

Normally, the data lies within a specified range. Consider that, a person is working in a company
and drawing a salary of ₹70,000 per month. If you observe a transaction of ₹10 crore in his
account, this transaction would be an anomaly. This could be the result of a fraud or a cyber-
attack. So, these kinds of anomalies can be detected with the help of data science. The banking
software keeps track of your regular transactions and if something abnormal happens, it notifies
the customer. An unexpected change in the data pattern indicates that there is something

abnormal.
Some more examples of anomaly detection are:

• Is the email normal or spam?
• Is the seller genuine or not?
• Is the reading of the temperature normal?

To answer these types of questions, anomaly detection algorithms are used.

108