prediktion

Prediktion is the process of forming a statement about a future event or outcome based on evidence, models, or reasoning. It is used across disciplines to estimate what is likely to happen, sometimes with a quantified level of uncertainty.

The term stems from Latin predictio and is used in Swedish, Norwegian, and Danish with overlap to prophecy in religious contexts. In everyday language prediktion can refer to both scientific predictions and prophecies, while in technical contexts it often denotes a modeled estimate with associated uncertainty.

In statistics and data science, prediktion involves using a model to estimate a future observation or the distribution of possible values, rather than merely inferring a fixed parameter. Common methods include regression, time series analysis, and a broad range of machine learning algorithms. Predictions can be point estimates or probabilistic, giving predictive intervals. Forecasts are a form of prediktion focused on a time horizon, such as weather or economic trends.

Evaluation of prediktion uses metrics such as RMSE, MAE, calibration, and scoring rules like the Brier score, depending on the task. Reliability depends on data quality, model validity, and assumptions. Limitations include uncertainty, non-stationarity, overfitting, and ethical concerns about bias, transparency, and the consequences of incorrect predictions.

Examples include weather prediktion, stock price predictions, epidemiological forecasts, and demand projections in operations research.