Predicting filters are filters that estimate the next value in a time series from the known previous values. In contrast to smoothing filters, in predicting filters the output at the moment depends only on the values at preceding moments: 
.
Predicting filters are used in time series analysis. In finance, for example, predicting filters are used to forecast stock prices, currency exchange rates, and other economic indicators from the known values of these indicators in the past.
Examples of predicting filters are the exponential filter and Kalman filter.
The complementary concept is smoothing filter, that may use input values for time moments both preceding and following the current time and, hence, cannot be used in the real time, only post-factum – when the whole record is already available.