There are various classification algorithms. The most common and simple example, one that anyone has to refer to if they want to know more about classification algorithms, is the Iris dataset; a dataset on flowers. Researchers constantly use this example in their research papers. In the video below, you will see that there are 150 observations, 150 rows and there are 5 columns. First four columns are sepal length, sepal width, petal length and petal width; these are independent variables. The last column is class, there are three classes in these flowers- Setosa, Versicolor and Virginica; this is the dependent variable. Now, we want to predict what should be the class of a particular flower that has been input as data; for this we use the classification algorithm. If you have data of 150 observations and new data that is previously unseen by the system is introduced, the system must be able to predict to which class the new data belongs to. So, you can see there is learning here. After a given number of observations, when the algorithm has been fully “trained”, it can make accurate predictions by itself. We can also take the example of hospitals managing patient data. In the patient database, we can have three classes; e.g. class “1A” patient is normal, class “2A” patient is suspected to have a disease and class “3A” patient is likely to die. If we are considering tests from one to n number of deaths, based on the existing data the patient belongs to either of the three classes. The model learns based on the existing data; and when a new patient is admitted, the system must apply the “trained” algorithm and easily classify. It can be used in the banking sector as well. The algorithm can be trained with existing data of customers, this data will teach the system to classify customers according to certain conditions and characteristics of behaviour and transactions. When a new customer data is added, the system must be able to accurately predict which class the customer belongs to and classify him/her accordingly.