Fuzzy Time Series Forecasting Using High-Order Interval Ratios and Frequency Density

E. Sarwoko,Etna Vianita,Adi Wibowo
DOI: https://doi.org/10.1109/ICICoS62600.2024.10636928
2024-07-17
Abstract:Advancements in fuzzy time series (FTS) forecasting in recent times have shown that combining interval ratio and frequency density in high-order approaches may lead to better prediction accuracy. By using a combined strategy of frequency density adjustments and interval ratio-based partitioning, the new FTS model for high-order presented in this research greatly outperforms the current models. To better capture the inherent variability in the data, researcher suggest an improved approach that divides the universe of discourse into interval ratios that were then further subdivided into tighter sub-intervals according to their frequency density. The model may also take into account longer temporal connections to the use of high-order fuzzy logic linkages, which is essential for complicated time series data like financial markets. A thorough case study on exchange rate in daily of USD to IDR from January-May 2020 forecasting was used to evaluate our methodology, and the results demonstrate a significant increase in prediction accuracy and error measures over earlier models. The results showed that our model was resilient in managing the nonlinear and dynamic character of the data, as seen by decreased Mean Squared Error (MSE) and Average Forecasting Error Rate (AFER). This work had practical implications for economic forecasting and other fields where accurate prediction was crucial, in addition to contributing to the theoretical development of FTS.
Mathematics,Computer Science,Engineering
What problem does this paper attempt to address?