Types of Data

Each column(feature) in a training set has a datatype associated with it. Those datatypes can be grouped into numerical, categorical, and date and(or) time.

Most ML algorithms perform well only with numerical data, so categorical and datetime values need to be converted into numbers.

For date and time, features such as 'year', 'month', and similar can be extracted, depending on the task. These are already numerical values, so they can be used directly.

Categorical data is a little more challenging to deal with.

Types of Categorical Data

Categorical data is classified into two types:

• Ordinal data is a type of categorical data in which categories follow a natural order. For example, level of education (from elementary school to Ph.D.) or rates (from very bad to very well), etc.;

• Nominal data is a type of categorical data that follows no natural order. For example, name, gender, country of origin, etc.

Converting ordinal and nominal data types into numerical values requires different approaches, so they must be handled separately.