Akademik Veri Yönetim Sistemi
Turkish Journal of Engineering and Environmental Sciences, cilt.23, sa.3, ss.173-180, 1999 (Scopus)
Classical time series analysis requires many assumptions such as the normality of data, linearity in the autocorelation coefficient and statistical parameter estimations. It is almost impossible to find all these assumptions applicable in stochastic time series generation or simulation. This paper provides a simple fuzzy-probabilistic method for the time series analysis. The basis of the methodology is to construct the fuzzy base rule domain from the available daily maximum temperature records at Kandilli observatory in Istanbul. The new concepts of transition and cumulative probability procedures are employed for taking decision among the alternative consequent fuzzy sets prior to the defuzzification. Classical time series analysis requires many assumptions such as the normality of data, linearity in the autocorelation coefficient and statistical parameter estimations. It is almost impossible to find all these assumptions applicable in stochastic time series generation or simulation. This paper provides a simple fuzzy-probabilistic method for the time series analysis. The basis of the methodology is to construct the fuzzy base rule domain from the available daily maximum temperature records at Kandilli observatory in Istanbul. The new concepts of transition and cumulative probability procedures are employed for taking decision among the alternative consequent fuzzy sets prior to the defuzzification.