What is Forecasting?
Forecasting is a type of predictive modeling that estimates future values or outcomes using historical data. It is widely used across fields such as business, healthcare, finance, and operations to support planning and decision-making. Unlike other types of predictive modeling that focus on classifying or grouping data, forecasting aims to predict how much or how many of something will occur at a future time.
Many forecasting models come from supervised learning, where an algorithm is trained to predict a continuous outcome based on a set of input features. These models can incorporate a wide range of information to generate forecasts such as economic indicators, customer characteristics, environmental variables, or past behaviors. Examples include predicting customer demand, estimating future revenue, or projecting resource needs.
Forecasting Methods
Common forecasting methods include the following types.
Linear regression and related models are often used when the relationship between predictors and outcomes is expected to be linear and interpretable. They’re useful for quick, transparent forecasts and are easy to explain to non-technical audiences. Nonlinear regression models are also available, but they are more complex to specify and interpret, often requiring deeper subject-matter or modeling expertise.
Tree-based models like decision trees, random forests, and gradient boosting machines are powerful for capturing nonlinear relationships and interactions without requiring extensive data preprocessing. They perform well when the dataset includes a mix of numeric and categorical variables.
Neural networks are well-suited for complex forecasting tasks with large datasets and many features, especially when the relationships between inputs and outputs are highly nonlinear or layered.
Time series methods such as ARIMA and exponential smoothing are specifically designed for sequential data are most appropriate when the goal is to model temporal patterns like trends or seasonality directly.
Evaluation
Forecasting models are typically evaluated using performance metrics such as:
Effective forecasting depends on the quality of the input data, thoughtful feature selection, and validation using held-out or future observations.
« Back to Glossary Index