A new method for temperature prediction and the TAIFEX Forecasting based on fuzzy logical relationship and double Interval division

This paper proposes a new method in time series forecasting for the daily temperature data set and the TAIFEX series (1996), with a novel approach in interval setting and model fuzzification. This model utilizes intervals with overlap which assigns each one of the individual entity of the set a weighted fuzzy counterpart. Fuzzified corresponding to each datum is a linear combination of two successive linguistic variables. The weights in the linear combination are degrees of memberships by which the datum belongs to two successive fuzzy sets. The second degree is the complement of the first one and vice versa. Fuzzy Logical Relation (FLR) is used to cluster the set and the mean method to defuzzification.