Forecasting the future price of pistachio in agricultural commodity exchange using of the hybrid model of Wavelet-XGBoost

In recent years, the Iran Commodity Market has always been associated with price destabilizing fluctuations. Thus, it is need to use machine Learnings in the form of forecasting methods to recognize early events and prevent damage caused by these fluctuations. Considering the important position of the pistachio trading ring in the Iran Commodity and also the need to use appropriate tools to correctly diagnosis the future price, The purpose of this study is to design and build a suitable hybrid model based on XGBoost and compare its performance with other machine learning models in order to accurately forecast the future price of pistachio. In this study, software’s of Matlab, Rapid Miner and Scikit learn were used to build the proposed hybrid model. The results of applying the wavelet theory showed that the error value of price data was reduced and the data had a stable trend (white noise). Also, the results of the Performance of Auto-Encoder network and the Genetic algorithm showed that the optimal lag of one is the best input variable for forecasting the future price of pistachios in the period under review. Based on goodness of fit indices, the proposed model of this study, Wavelet-XGBoost in comparing to other data mining models, had a better performance in forecasting the future price of pistachios. Also, out-of-sample forecasting with the selected model showed that the forecasted new prices have little difference with the real data, which indicates the efficiency and accuracy of the selected hybrid model. According to the obtained results, it is strongly recommended to use the proposed model based on XGBoost algorithm to forecast the price of other agricultural products.