Forecasting Direction of Trend of a Group of Analogous Time Series Using F-Transform and Fuzzy Natural Logic

Abstract

We present an idea to group time series according to similarity of their local trends and to predict future direction of the trend of all of them on the basis of forecast of only one representative. First, we assign to each time series an adjoint one, which consists of a sequence of the F1-transform components. Then, they are grouped together according to their similarity, a principal time series is selected in each group and its future course is forecasted. Finally, directions of trends of the other time series from the same group are computed using special methods of fuzzy natural logic.

Keywords:

• F-transform
• time series
• Fuzzy Natural Logic
• evaluative linguistic expressions