A statistical approach to data analytics that attempts to determine the relationship between a set of independent variables.
What is it?
Regression is a statistical approach to modeling data. There are a number of different types of regression models — linear, polynomial, non-linear etc. — which attempt to define patterns between different variables.
It can be useful for identifying a pattern in the data that allows you to spot trends. For instance, a sales manager trying to predict next month’s figures might use regression analysis to work out, using previous sales data, the likely impact of the summer weather.
What’s in for you?
Regression analysis is one of many traditional statistical methods, which are well understood and relatively inexpensive, that can be used to make predictions.
What are the trade offs?
Essentially, regression works by using data to plot a graph and trying to fit a line that describes the relationship between the variables. In many cases, the behavior being modeled can be more complex than your fitted line suggests. So regression may be too simplistic for the behavior you’re trying to predict.
If your model is wrong, you may need a more sophisticated approach. You’ll essentially need to train more complex models.
How is it being used?
Regression is best suited to relatively simple prediction problems such as demand forecasting, or pricing. If your business can be run from a spreadsheet, regression might well be a good fit for you.