0
$\begingroup$

I use the two terms as follows:

A prediction model gets features (which can be a time series) as input and gives a fixed-length output (might be multiple values, but "atomic" in some sense)

Examples:

  • Dog vs Cat
  • Spam vs No Spam
  • Weather for tomorrow
  • Sales of next month

A forecasting model gets a time series as input and outputs a single time series.

Examples:

  • Amount of airline passengers: You can predict the value for the next month, but a forecast can be over an arbitrary time frame (with deteriating quality, of course)
  • Weather
  • Sales

I've never seen textbook definitions of those terms. Do you know any source that compares the two terms?

Bonus points if you give me a good German translation for both terms. By now, I call both "Vorhersage" - I have the impression there is no other German word for it.

$\endgroup$
  • $\begingroup$ Simply stated, forecasting is prediction using time series data. $\endgroup$ – Ankit Seth Jun 28 '18 at 9:49
  • $\begingroup$ @AnkitSeth This is what Daniel Chepenko says. $\endgroup$ – Martin Thoma Jun 28 '18 at 11:08
1
$\begingroup$

Prediction is more general term. It can be defined as a an expectation for a combination of different predictors (random variables).

Forecasting would be a subset of prediction and mostly related to time-series. Any time you predict into the future it is a forecast. All forecasts are predictions, but not all predictions are forecasts.

Moreover, forecasting always have error, that is why we have confidence interval of 5%, 2.5%, 1% and residuals for modeling. There is no such thing as so called "perfect forecasting", but we can maximize the model accuracy.

On contrast, prediction sometime can be perfect. I also believe that model should follow data not the other way around.

| improve this answer | |
$\endgroup$
0
$\begingroup$

Forecast is more probabilistic events and is used for more certain events. Example of predictions could be predicting future stock prices, house prices, percentage of house owners buying home insurance etc. Each model is made up of a number of predictors (variables) and based on predictors we can create a predictive model. Predictions relies on methodologies like Logistic Regression (predict values based on prior observation of the data set), Decision trees (prediction of every possible outcome of a decision), Time series analysis (related to time intervals).

Prediction is utilized more in business and economics while forecasting takes place in weather and earthquakes and it is always associated with a time dimension in the future. Prediction is the subset of Forecasting. For every forecast we will do one or more prediction.

You can read more here.

| improve this answer | |
$\endgroup$
0
$\begingroup$

I guess the same as between Water and Dihydroxy-oxygen. One is in the common language, the other refers to a scientific discipline.

  • Forecasts are qualitative, prediction may (or may not) be qualitative.

  • Forecasts are based on an explicit model, prediction may (or may not) be based on intuition or medium powers.

  • Forecasts are made by fitting the parameters of the model, prediction may (or may not) use gestimate or bestimate for the parameters.

  • Forecasts are corrected for seasonality and other exogenous factors, prediction may (or may not) indicte a trend.

  • Forecasts errors and computed a priori at the time of the forecast, prediction may (or may not) reveal to be right.

  • Forecast use algorithms, prediction may (or may not) use crystal balls, cards, expert (or not expert) judgment.

Not that predictions are necessarily bad. Simply that all forecasts are predictions.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.