Chaotic Crude Oil Time-Series Prediction Using Multi-Objective Artificial Vulture Optimization Algorithm-Based Bidirectional Gated Recurrent Unit

Crude oil is a resource of non-renewable liquid petroleum in existence by nature which is extracted from the earth and evolved into various products. However, accurately predicting crude oil prices is difficult due to non-stationarity and market noise. In this research, the Multi-Objective Artificial Vulture Optimization Algorithm-based Bidirectional Gated Recurrent Unit (MOAVOA-BGRU) to accurately predict the chaotic crude oil time-series price using Deep Learning (DL). Initially, the data is gathered from West Texas Intermediate (WTI) dataset and min-max normalization is performed to enhance the model performance. The technical indicators approach is used to extract the features and MOAVOA technique is established to select the appropriate features. At last, BGU is utilized to predict the crude oil price cost. The existing techniques like Variational Mode Decomposition-Recurrence Plot-BGU (VMD-RP-BGRU), Empirical Mode Decomposition-Autoregressive Integrated Moving Average-based Long Short-Term Memory (EEMD-ARIMA-LSTM), and Local Mean Decomposition-Stochastic and Deterministic-ARIMA-LSTM (LMD-SD-ARIMA-LSTM) are compared with proposed technique. The MOAVOA-BGRU achieves better Mean Average Error of 1.024, 1.0207, and 0.012 in terms of weekly, daily, and monthly spot prices compared to VMD-RP-BGRU, EEMD-ARIMA-LSTM, and LMD-SD-ARIMA-LSTM respectively.