Simultaneous Previews for Time Series Forecasting

Time series data consist of sequences of real numbers, representing the measurements or observations of a real variable at equal time intervals. Huge amounts of time series data are available today in many domains, like auctions, new stock offerings, industrial processes. People often desire estimates of the future behavior of partial time series. In this paper we present a data driven forecasting method, that we call Similarity-Based Forecasting (SBF). The forecast is displayed graphically in a forecasting preview interface, allowing users to analyze the effects of alternative pattern matching parameters.