Variational Mode Decomposition based Machine Learning Models Optimized with Genetic Algorithm for Price Forecasting

Abstract

Accurate and timely price information and forecasting help in making efficient plans and strategies. Non-linearity and non-stationarity behaviour of price data create problems in price forecasting. In this paper, variational mode decomposition (VMD) based optimised genetic algorithm (GA) hybrid machine learning (ML) models have been proposed. The VMD algorithm is employed to decompose the price data into intrinsic mode functions (IMFs) which is further forecasted using ML models namely support vector regression (SVR) and random forest (RF). The practical use of the SVR and RF models is limited because the accuracy of ML models heavily depends on a proper setting of hyper-parameters. Therefore, these model hyper-parameters are optimized using GA. Further, the forecasted values of IMFs through the GA optimised SVR and RF are aggregated for the final forecast. The results of the proposed model are benchmarked with the comparative models. The proposed VMD-GA-RF and VMD-GA-SVR models are tested on the weekly onion price of the Delhi and Nashik market. The results clearly demonstrate that the combination of VMD and GA optimized models can improve the performance of the prediction of the dataset.