Exponential smoothing is a form of time series forecasting that uses a weighted average of past observations to estimate current and future values. It can be used for forecasting non-seasonal data, such as sales or inventories, or for seasonal data, such as climate or campaign performance.
Exponential smoothing works by progressively adjusting its estimates of future values based on the most recent historical data. It assigns a larger weight to newer data points, and a smaller weight to more dated data points. By doing this, the exponential smoothing technique is able to capture recent trends in the data while reducing the amount of noise that is usually present in data points that are further away from the current time.
In contrast to other forecasting techniques, such as autoregressive moving average (ARMA) models, exponential smoothing is relatively simple to understand and implement. Its simplicity makes it a useful tool for forecasting short-term data points in computer models.
Exponential smoothing is also known as exponentially weighted moving average (EWMA).
