A prediction () or forecast is a statement about a future event or about future data. Predictions are often, but not always, based upon experience or knowledge of forecasters. There is no universal agreement about the exact difference between "prediction" and "estimation"; different authors and disciplines ascribe different connotations.
Future events are necessarily uncertain, so guaranteed accurate information about the future is impossible. Prediction can be useful to assist in making plans about possible developments.
Opinion In a non-statistical sense, the term "prediction" is often used to refer to an informed guess or opinion.
A prediction of this kind might be informed by a predicting person's abductive reasoning, inductive reasoning, deductive reasoning, and experience; and may be useful—if the predicting person is a knowledgeable person in the field.
The Delphi method is a technique for eliciting such expert-judgement-based predictions in a controlled way. This type of prediction might be perceived as consistent with statistical techniques in the sense that, at minimum, the "data" being used is the predicting expert's cognitive experiences forming an intuitive "probability curve."
Statistics In statistics, prediction is a part of statistical inference. One particular approach to such inference is known as predictive inference, but the prediction can be undertaken within any of the several approaches to statistical inference. Indeed, one possible description of statistics is that it provides a means of transferring knowledge about a sample of a population to the whole population, and to other related populations, which is not necessarily the same as prediction over time. Forecasting usually requires time series methods, while prediction is often performed on cross-sectional data.
Statistical techniques used for prediction include regression and its various sub-categories such as linear regression, generalized linear models (logistic regression, Poisson regression, Probit regression), etc. In case of forecasting, autoregressive moving average models and vector autoregression models can be utilized. When these and/or related, generalized set of regression or machine learning methods are deployed in commercial usage, the field is known as predictive analytics.
In many applications, such as time series analysis, it is possible to estimate the models that generate the observations. If models can be expressed as transfer functions or in terms of state-space parameters then smoothed, filtered and predicted data estimates can be calculated. If the underlying generating models are linear then a minimum-variance Kalman filter and a minimum-variance smoother may be used to recover data of interest from noisy measurements. These techniques rely on one-step-ahead predictors (which minimise the variance of the prediction error). When the generating models are nonlinear then stepwise linearizations may be applied within Extended Kalman Filter and smoother recursions. However, in nonlinear cases, optimum minimum-variance performance guarantees no longer apply.
To use regression analysis for prediction, data are collected on the variable that is to be predicted, called the dependent variable or response variable, and on one or more variables whose values are hypothesized to influence it, called independent variables or explanatory variables. A functional form, often linear, is hypothesized for the postulated causal relationship, and the parameters of the function are estimated from the data—that is, are chosen so as to optimize is some way the fit of the function, thus parameterized, to the data. That is the estimation step. For the prediction step, explanatory variable values that are deemed relevant to future (or current but not yet observed) values of the dependent variable are input to the parameterized function to generate predictions for the dependent variable.
Machine learning and artificial intelligence In recent decades, prediction has become a central task in machine learning and artificial intelligence research. Supervised learning algorithms, such as support vector machines, decision trees, and neural networks, are trained on historical datasets to predict outcomes on new, unseen data. These models are widely applied in domains such as natural language processing, computer vision, health informatics, and financial technology. Recent studies have emphasized the importance of model interpretability and fairness, since predictions can influence critical decisions in healthcare, criminal justice, and public policy.
An unbiased performance estimate of a model can be obtained on hold-out test sets. The predictions can visually be compared to the ground truth in a parity plot.
Science
In science, a prediction is a rigorous, often quantitative, statement, forecasting what would be observed under specific conditions; for example, according to theories of gravity, if an apple fell from a tree it would be seen to move towards the center of the Earth with a specified and constant acceleration. The scientific method is built on testing statements that are logical consequences of scientific theories. This is done through repeatable experiments or observational studies.
A scientific theory whose predictions are contradicted by observations and evidence will be rejected. New theories that generate many new predictions can more easily be supported or falsified (see predictive power). Notions that make no testable predictions are usually considered not to be part of science (protoscience or nescience) until testable predictions can be made.
Mathematical equations and models, and computer models, are frequently used to describe the past and future behaviour of a process within the boundaries of that model. In some cases the probability of an outcome, rather than a specific outcome, can be predicted, for example in much of quantum physics.
In microprocessors, branch prediction permits avoidance of pipeline emptying at branch instructions.
In engineering, possible failure modes are predicted and avoided by correcting the failure mechanism causing the failure.
Accurate prediction and forecasting are very difficult in some areas, such as natural disasters, pandemics, demography, population dynamics and meteorology. For example, it is possible to predict the occurrence of solar cycles, but their exact timing and magnitude is much more difficult (see picture to right).
In materials engineering it is also possible to predict the life time of a material with a mathematical model.
In medical science predictive and prognostic biomarkers can be used to predict patient outcomes in response to various treatment or the probability of a clinical event.
Hypothesis Established science makes useful predictions which are often extremely reliable and accurate; for example, eclipses are routinely predicted.
New theories make predictions which allow them to be disproved by reality. For example, predicting the structure of crystals at the atomic level is a current research challenge. In the early 20th century the scientific consensus was that there existed an absolute frame of reference, which was given the name luminiferous ether. The existence of this absolute frame was deemed necessary for consistency with the established idea that the speed of light is constant. The famous Michelson–Morley experiment demonstrated that predictions deduced from this concept were not borne out in reality, thus disproving the theory of an absolute frame of reference. The special theory of relativity was proposed by Einstein as an explanation for the seeming inconsistency between the constancy of the speed of light and the non-existence of a special, preferred or absolute frame of reference.
Albert Einstein's theory of general relativity could not easily be tested as it did not produce any effects observable on a terrestrial scale. However, as one of the first tests of general relativity, the theory predicted that large masses such as stars would bend light, in contradiction to accepted theory; this was observed in a 1919 eclipse.
Medicine and healthcare Predictive medicine
Prognosis
Clinical prediction rules
Finance
Mathematical models of stock market behaviour (and economic behaviour in general) are also unreliable in predicting future behaviour. Among other reasons, this is because economic events may span several years, and the world is changing over a similar time frame, thus invalidating the relevance of past observations to the present. Thus there are an extremely small number (of the order of 1) of relevant past data points from which to project the future. In addition, it is generally believed that stock market prices already take into account all the information available to predict the future, and subsequent movements must therefore be the result of unforeseen events. Consequently, it is extremely difficult for a stock investor to anticipate or predict a stock market boom, or a stock market crash. In contrast to predicting the actual stock return, forecasting of broad economic trends tends to have better accuracy. Such analysis is provided by both non-profit groups as well as by for-profit private institutions.
Some correlation has been seen between actual stock market movements and prediction data from large groups in surveys and prediction games.
An actuary uses actuarial science to assess and predict future business risk, such that the risk(s) can be mitigated. For example, in insurance an actuary would use a life table (which incorporates the historical experience of mortality rates and sometimes an estimate of future trends) to project life expectancy.
Sports
Predicting the outcome of sporting events is a business which has grown in popularity in recent years. Handicappers predict the outcome of games using a variety of mathematical formulas, simulation models or qualitative analysis. Early, well known sports bettors, such as Jimmy the Greek, were believed to have access to information that gave them an edge. Information ranged from personal issues, such as gambling or drinking to undisclosed injuries; anything that may affect the performance of a player on the field.
Recent times have changed the way sports are predicted. Predictions now typically consist of two distinct approaches: Situational plays and statistical based models. Situational plays are much more difficult to measure because they usually involve the motivation of a team. Dan Gordon, noted handicapper, wrote "Without an emotional edge in a game in addition to value in a line, I won't put my money on it". These types of plays consist of: Betting on the home underdog, betting against Monday Night winners if they are a favorite next week, betting the underdog in "look ahead" games etc. As situational plays become more widely known they become less useful because they will impact the way the line is set.