Descriptive analytics are great and allow us to construct a snapshot of where we’ve been and how we’re doing now. However, to really get a strategic advantage, wouldn’t it be great to be able to see into the future and see what our business intelligence picture is going to look like before it happens? Welcome to the world of predictive analytics.

## The basics

Predictive analytics use mathematical and statistical techniques to identify patterns in data in order to make an estimate of unknown events. These events might be unknown because they are due to happen in the future, or they may have already happened but they weren’t recorded. Either way, the important concept is that patterns are identified in data in order to build a model to predict what the other, unknown, events are most likely to be.

## What does it look like?

Probably the best known example of predictive analytics is the weather forecast. Predicting the weather is an incredibly complicated task, but in simple terms the forecasters take measurements, from a great number of locations, of things like air pressure, wind speed and humidity, then analyse this information to predict what the weather will be like in a few hours, days or weeks time.

“By artificially continuing patterns in #data, a #predictiveanalytics model can estimate what is on the…

This works because there are underlying systems which cause the weather, so by taking measurements of things such as air temperature, forecasters can create a model of this system. By rolling this model forwards, continuing the patterns seen so far, the model can give a reasonable estimate of what is on the horizon.

## Cheat sheet

With predictive analytics involving some advanced mathematical techniques, there are some terms you may come across which belong in a statistics lecture. Here, we give you a quick reference guide to some common terms (note: examples are entirely made up).

Model | A system described mathematically. For example, “Humidity” – “Temperature” + “It always rains in the UK” = “Chance of rain today” |

Linear Regression | A type of model used to predict a specific value. Linear regression takes a number of measurements and combines them into an equation to predict an unknown measurement. For example, “(Humidity x 3) + (Wind x 5) = Size of storm landing tomorrow” |

Classification | A common predictive task where the aim is to predict which category a certain input falls into. For example, taking all of the measurements, weather forecasters have access to, into account and classifying tomorrow’s weather according to the “Sun”, “Rain”, “Windy” symbols we see on the TV. |

Predictor | A value that can be measured to predict something else. For example “humidity” may be a predictor of “chance of rain”. |

Dependent variable | A value to be predicted based on predictors, “Chance of rain” in the example above. |