Before trying to predict some observed phenomena given some data, we need to first "clean/modify" the data so that one can apply predictive algorithms to said data. This process is called feature engineering.

​

Examples of practical applications where feature engineering is used often include: customer retention (churn rate analysis) or fraud detection (outlier detection).

​

For some impractical uses of data, see my research publications.

​

Lets start with some definitions:

​