A Modified of The Generalized Fuzzy Logical Relationship Method With High Order Fuzzy Time Series Based on Frequency Density Partition Nina Munazilla (a*), Farikhin (b), Sunarsih (b)
a) Faculty of Science and Mathematics, Diponegoro University
Jl. Prof. Soedarto, SH, Tembalang Semarang - 50275, Jawa Tengah, Indonesia
*fsm[at]undip.ac.id
Abstract
One of step in fuzzy time series for forecasting is to build fuzzy logic relationship. Many investigators have generalized a fuzzy logic relationship. Algorithms, computational methods and grouping the Fuzzy Logical Relationship (FLR) are three methods based on advanced to build a high order fuzzy time series models. The last kind model is used to determine the forecasting decisions. To improve the fuzzy time series of the approximate model, this paper presents a high-order fuzzy time series model denoted as GTS(M.N) based on the basis of generalization fuzzy logical relationship by proposing improvements of universe of discourse, historical data variations, partition stages and weighting stages. The first, define the universe of discourse, then by using historical data variations obtained the number of intervals. Second, the primary interval is partitioned based on frequency density into several sub-intervals. Third, perform different weightings on FLR to calculate the final forecasting value. The proposed model will be implemented in time series data of coffee production in 2000-2019. Based on these our experiment we have the resulted errors value is better the existing method.
Keywords: Fuzzy sets, Fuzzy time series, Frequency density, Partitioning
