A predictive model is a mathematical or statistical model used in artificial intelligence and machine learning to predict the value of a variable of interest based on historical data and observed patterns. Usually the event to be predicted is future, however predictive modeling can be applied to unknown events regardless of time.

Predictive models are used in a variety of applications, such as risk analysis, fraud detection, demand prediction, image classification, among others. These models are trained using a historical data set containing information about the relevant variables and the corresponding outputs.

Once trained, the predictive model can be used to make predictions about new observations. To do this, the model is provided with an input containing values of the input variables, and the model uses these values to make a prediction about the corresponding output value.

The predictive model thus gives rise to forecasts which, however, should be considered as probabilities and not as certain predictions, which will necessarily materialize. The plausibility of the results of the predictive models must be reported statistically. Then, their probability is conceivable depending on the size of the data set studied. Thus, the larger the number of data analyzed, the more the results of prognostic models can be considered as conceivable and accurate results.

Predictive models are built using supervised learning techniques, which require the existence of a labeled data set for model training. The accuracy of a predictive model is evaluated using error metrics such as precision, mean square error or ROC curve.