Data Intelligence, Business Analytics
Can anyone help me understand the difference between Prediction (as in Predictive Modeling) and Forecasting?
If forecasts can be obtained using Regression, then how do we differentiate between Forecasting and Prediction?Even though my example is specific I would like to know the difference in a generic sense.
Forecasting would be a subset of prediction. Any time you predict into the future it is a forecast. All forecasts are predictions, but not all predictions are forecasts, as when you would use regression to explain the relationship between two variables.
In book D. E. Catlin "estimation, control, and the discrete Kalman filter" the word prediction is used as:
hatx(k | j) for j=k is called the filtered estimate of x(k)
hatx(k | j) for j<k is called the predicted estimate of x(k)
hatx(k | j) for j>k is called the smoothed estimate of x(k)
Maybe it is possible to do something similar for the word forecast to see their different usages.
I'm president of Strategic Affairs Forecasting LLC and am a futurist that has made very careful distinctions between prediction and forecast for many years. Here are the key elements in my opinion:
-For a number of reasons many people do not believe that you can predict the future. I and other futurist argue that we are not about predicting the future, but about exploring it to improve decision-making today. I and most other futurists will say that a good forecast set is likely to contain several elements of the most possible futures relevent to the client's question.
-Predictions are spot/one-off estimates of a specific event in the future; usually at a specific point in time. The usual examples are gambling: Who will win the World Series this year? Kentuky Derby? etc. On occasion, folks will deliver a set of predictions and call them a forecast (e.g., 7 day weather forecasts)
-Forecasts are a set of possible futures that include probability estimates of occuring. The entire set of scenarios should have a probability close to 1. Often, forecasts have more generalized time points (e.g., next five years, next decade, mid-range future, etc). It is possible that a forecast group can consist of a set of predictions with associated probabilities.
-Most projects are about predictions. The user of the information wants an estimate of a specific future outcome at a specific time.
-Forecasts are more useful in strategic discussions about what are the useful variables/factors and what potential outcomes may exist. On occassion, the analyst may need to help the client disern whether a prediction or forecast is the most useful approach to their info needs.
-As a rule of thumb, predictions are most often outcome focused (who wins, sales, profit, etc). Forecasts tend to focus on process factors (what factors may influence the future and what are their potential outcomes).
Two further observations:
-Data mining focuses on identifying the factors associated with specific outcomes. I often relate data mining to sports handicapping. The goal is to identify the factors that have associated outcomes and then apply this knowledge to a specific situation in the future.
-The term 'predictive analytics' should be dropped in the long run. Most people do not believe that you can predict the future. Thus, this term is associated with ideas that folks do not believe...bad marketing.
Forecasting is all about numbers ...e.g how many ice cream cones to be manufactured in may, jun, july & so on.
prediction is all about bahaviroal ....which flavors are liked by customers most...which age group is likely to buy vanila, straberry etc.how will customers behave at morning, afternoon & evening at the ice cream shop etc.
Prediction is the generalize term & it's independent of time. Forecasting is the prediction with time as a one of the dependent variable. Eg-
Prediction- Predicting amount spend by user for certain case. It's happen over the period of time but not exact.
Forecasting- Best example is weather forecasting. Weather prediction is called forecasting since it's predicted with time as a dependent variable either for quarter or next month etc.