Prediction of Turkey Wheat Yield by Wavelet Fuzzy Time Series and Gray Prediction Methods

Abstract

This study presents the analysis of the wheat yield values for Turkey during the years 1946-2018. The 73-year wheat yield values (kg da-1) were analyzed by two different mathematical methods frequently used in the literature. Fuzzy Time Series (FTS) and Gray Prediction (GP) algorithms used in the time series field are the two analysis methods used in this study. In addition to these models, a hybrid model was created by the FTS method using the discrete wavelet transform (DWT) technique, which is one of the time series pre-processing methods. During the analysis, 53 annual wheat yield values were used to train the models, while 20 annual wheat yield values were used as the test set. Model performances were evaluated with mean square error (MSE) and coefficient of efficiency (CE) success criteria. The results revealed that the wavelet fuzzy time series (WFTS) models had the highest CE value and the lowest MSE value. Thanks to this method, which is used for the first time in the field of agricultural sciences, it is predicted to make more accurate decisions. Besides, it is thought that the hybrid model created with the study will shed light on other researchers for further developments.

Keywords

• wheat
• yield
• fuzzy time series
• wavelet transform
• grey prediction