Machine learning is the term coined for the process of a machine learning to ‘do things’ intelligently from the data provided. With historical data, the output variable is represented as a combination of input variables. The prediction or output values for a set of input values can be numerical or categorical. When the values are continuous, regression is used. Regression is used extensively in finance, pricing, investing etc. So in conclusion, Regression involves the calculation of relationship between output and multiple input variables.

Types of Regression:

Linear Regression: Single input variable is used to predict output variables. The relationship between the input variable and output is more or less a single line that fits the data well. y= a+ b*x ; y is output, x is input. A variation of this is multiple linear regression where more than one input variable is used to predict output.

Ridge Regression: Variation of multiple linear regression. It uses an estimator different from ordinary least squared error, which causes shrinkage in coefficients. The estimator has lower variance and is biased.

Lasso Regression: Least Absolute Shrinkage and selection operator. It penalizes the sum of the coefficients to be less than a fixed value, thereby a few of the coefficients are also omitted resulting in feature selection.

Non-Linear Regression: The relationship between the input and output variables is not a linear one. It is a sum of products of variables. Y = a + bX^2 ; Y is output, X is the input

Logistic Regression: The output variable is a categorical variable.