Abstract
Fuzzy trend models have been proposed to represent the trend in time-series.A model for multivariate time series is also provided. These models are based on fuzzification of the step function. Then a number of parameters may become large and the shape of trend becomes unnatural sometimes. In this study a generalized model is proposed for representing trend naturally by fewer parameters. An estimation method for the proposed fuzzy trend model is provided and its applicability is shown by simulation studies. A practical application to financial time series is also demonstrated.
