In this segment, we would like to define supervised learning.

It is also called classification and we will describe various methods.

These are the basic methods.

We start with Twitter data example.

In this dataset, there are 10 instances,

then there are three attributes.

And the last one is the class attribute.

So in this case,

we just imagine that we have collected data and for the three attributes: celebrity,

verified account and number of followers.

And the class level is something we would like to learn or predict.

This user is influential or not.

And this dataset is very important because

some for the classification we usually have this kind of

a format and we will have collected data and try to learn the patterns.

So for the supervised learning,

the process is like this.

We have a training set,

dataset, then we have a learning algorithm,

after we learn the model,

we will use the model for the testing part.

But in reality, we will just use the model for classification test.

However, we usually include the testing

because we would like to see how good the model is.

And in the early example,

when we try to apply the model,

we predict that any user

is influential or not and in this example here,

we will say after we fall our e-mail messages,

we would like to classify some messages as spam or not.

So, here for example we have these unlabelled instances we would like to know the class.

But m here is the model.

So the key is to find the mapping.

So we call the M the smart Mabi maps x 2 y,

y is the class.

So for supervised learning algorithms,

there are numerous number of and we won't be able to cover all of

them but we will try to cover those very basic and intuitive as well as popular ones.

There are three listed here.

So one category it's called the classification.

It's Decision Tree Learning and k-

Nearest Neighbor Classifier then another category is regression.

So well spend time on all of these three methods.

And another reason for us to choose or start with

these three methods because usually people use one of

them as their baseline for a performance comparison and also all these three methods are

intuitive and they are more understandable than other methods.

So we start with Decision Tree Learning.

So when we discuss Decision Tree,

we focus on induction or induce a tree.

So we still use an example to show you. What is a Decision Tree?

So this is the dataset we use in

the very beginning to illustrate the need for supervised learning.

So this is the one of the trees we can induce from this dataset.

And for the training data,

we have known classes and the long the tree if a tree like

this will be applied to those data or instances with unknown classes.

So we say unknown classes usually we mention we talk about this influential classes.

So the question here now is,

obviously I can have many trees and which attribute like we have three attributes,

which attribute should we start as the root note?