Welcome back. In this video we will be

learning about Predictive Modeling and Data Analytics.

Prediction is a fundamentally important activity and it can take many different forms.

The process of using data and models to

make predictions is often called predictive modeling.

In predictive modeling we would like to

understand the relationship between a set of variables.

The variable of interest on which we would like to make the prediction is called

the target variable or other variables

as in the target variables are called predictive variable.

You are likely to encounter many different ways to refer to those variables.

In this course we stick to the terminology introduced here.

A predictive model needs predict variables with a target variable.

That is the predictive model takes predictive variable as

input and produce an output which can be related to the target variable.

Why are we interested in predictive modeling?

As I've already mentioned we're often interested in making predictions.

A related and sometimes even more important goal

higher is to get inside the relationship among the different variables.

That is we would like to gain

some intuitive and accurate understanding of how different things are related.

Our personal experience will suggest that such knowledge is often quite valuable.

Predictive modeling can be roughly divided into two types: regression and classification.

The difference lies in the type of the part of the variable.

In classification the target variable is a binary or categorical.

Examples of categorical variable include the customer churn,

loan default, and to buy or not to buy our product.

Our customer churn occurs when the customers stops

purchasing their product or subscribing to a service.

Unknown is default if a customer fails to pay back unknown.

In all these cases there's a limited number of possible outcomes.

A customer either stays with a company or not,

unknown is either default or not,

and someone may choose to buy or not to buy a given product.

In contrast the target variable in reggression is continuous

which can theoretically have an infinite number of possible values.

Examples of continuous variables include house prices,

sales amount, and temperature.

Note that each of them can take many different values.

We'll start this model by learning how to

predict a continuous variable using linear regression.

Then, we'll discuss several important concepts in predictive modeling.

We will wrap up the module by practicing building

predictive models using a tool called XL Miner.