A HYBRID WAVELET-HIGH ORDER FUZZY COGNITIVE MAPS AND RANDOM FOREST REGRESSION APPROACH FOR TIME SERIES FORECASTING

Year 2025, Volume: 26 Issue: 3, 231 - 245, 25.09.2025

Aslı Kaya Karakütük

https://doi.org/10.18038/estubtda.1671595

Abstract

Time series forecasting becomes more critical, especially when linear, non-stationary and uncertain data are available. This paper proposes an innovative hybrid model that combines wavelet transforms, high-order fuzzy cognitive maps and random forest regression for high-accuracy prediction of time series. This hybrid approach aims to overcome the limitations of traditional time series analysis methods. In the approach, wavelet coefficients are enhanced by the integration of higher-order fuzzy cognitive maps, which allows for efficient modeling of nonlinear relationships through quadratic interactions. Four different wavelet transforms (Morlet, Mexican Hat, Haar, Daubechies) are used in the model and these structures are systematically compared. The model is trained with a Random Forest Regression model and hyperparameter optimization was performed using GridSearchCV. Model performance is evaluated on data sets of atmospheric CO2 concentrations, El Niño Sea surface temperatures and sunspot activity records from Mauna Loa Observatory. A comprehensive analysis is presented by evaluating the model performances with multiple metrics such as symmetric mean absolute percentage error, root mean squared error, mean absolute percentage error, and mean absolute scaled error. The experimental results show that the Mexican Hat wavelet performs the best in all data sets. It outperformed the other methods with root mean square error values of 0.2414 on CO2 data, 0.1621 root mean square error on El Niño temperature prediction and 4.3279 root mean square error on sunspot activity. While continuous wavelets are more successful than discrete wavelets, the integration of hybrid structure significantly improves the model's ability to capture nonlinear relationships. This research not only proves the superiority of wavelet-based hybrid models in time series analysis, but also demonstrates the potential for practical application in areas such as climate, meteorology, and space weather studies.

Keywords

Time series prediction , Wavelet , Transform , Fuzzy cognitive maps , Hybrid model , Random forest

References

• [1] Ozdemir O. The use of artificial neural networks in time series modeling and an application. MSc Thesis, Anadolu University, Institute of Science, Eskişehir, 2008.
• [2] Wei WWS. Time series analysis: univariate and multivariate methods. Pearson Education Inc., USA, 2006.
• [3] Bicen C. Comparison of forecasts made by Box-Jenkins time series analysis method and feedforward artificial neural networks. MSc Thesis, Hacettepe University, Institute of Health Sciences, Ankara, 2006.
• [4] Alrumaih MR, Al-Fawzan MA. Time series forecasting using wavelet denoising: an application to Saudi stock index. J King Saud Univ Eng Sci 2002; 14(2): 221-233.
• [5] Rioul O, Vetterli M. Wavelets and signal processing. IEEE Signal Process Mag 1991; 8(4): 14-38.
• [6] Lee G. Wavelets and wavelet estimation: a review. J Econ Theory Econometrics 1998; 4(1): 123-157.
• [7] Yang S, Liu J. Time-series forecasting based on high-order fuzzy cognitive maps and wavelet transform. IEEE Trans Fuzzy Syst 2018; 26(6): 3391-3402.
• [8] Labat D. Recent advances in wavelet analyses: part 1. a review of concepts. J Hydrol 2005; 314: 275-288.
• [9] Ofori-Ntow Jnr E, Ziggah YY, Relvas S. Hybrid ensemble intelligent model based on wavelet transform, swarm intelligence, and artificial neural network for electricity demand forecasting. Sustain Cities Soc 2021; 66: 102679.
• [10] Wu X, Zhou J, Yu H, Liu D, Xie K, Chen Y, Hu J, Sun H, Xing F. The development of a hybrid wavelet-ARIMA-LSTM model for precipitation amounts and drought analysis. Atmosphere 2021; 12(1): 74.
• [11] Ham YG, Kim JH, Luo JJ. Deep learning for multi-year ENSO forecasts. Nature 2019; 573: 568-572.
• [12] Ferkous K, Chellali F, Kouzou A, Bekkar B. Wavelet-Gaussian process regression model for forecasting daily solar radiation in the Saharan climate. Clean Energy 2021; 5(2): 316-328.
• [13] Karatzinis GD, Boutalis YS. A review study of fuzzy cognitive maps in engineering: applications, insights, and future directions. Eng 2025; 6(2): 37.
• [14] Statsmodels. El Niño Sea Surface Temperature dataset. Accessed July 3, 2025. https://www.statsmodels.org/dev/datasets/generated/elnino.html
• [15] Soltani S. On the use of the wavelet decomposition for time series prediction. Neurocomputing. 2002;48(1-4):267-277. doi:10.1016/S0925-2312(01)00648-8
• [16] Burrus CS, Gopinath RA, Guc H. Introduction to wavelets and wavelet transforms. Texas: Prentice Hall, 1998.
• [17] Goswami JC, Chan AK. Fundamentals of wavelets: theory, algorithm, and applications. John Wiley & Sons, USA, 1999.
• [18] Bahramian P, Saliminezhad A. Does capacity utilization predict inflation? wavelet-based evidence from United States. Comput Econ 2021; 58: 1103-1125.
• [19] Kosko B. Fuzzy cognitive maps. Int J Man-Machine Stud 1986; 24(1): 65-75.
• [20] Axelrod R, ed. Structure of decision: the cognitive maps of political elites. Princeton Legacy Library, 1976.
• [21] Stach W, Kurgan L, Pedrycz W. Higher-order fuzzy cognitive maps. In: Proc Annu Meet North Amer Fuzzy Inf Process Soc Conf, IEEE, 2006: 166-171.
• [22] Devhunter. Random Forest Algorithm. Devhunter Personal Blog, 20 Sep. 2018. Available from: https://devhunteryz.wordpress.com/2018/09/20/rastgele-ormanrandom-forest-algoritmasi/comment-page-1/
• [23] Breiman L. Random forests. Mach Learn 2001; 45(1): 5-32.
• [24] Dhanasekar C, Padmavathy C, Subramanian J. A computer-assisted crack predicting system for oil and gas pipelines using fuzzy cognitive map. Eur J Appl Sci 2015; 7(3): 145-151.