PREDICTING NIGERIA CRUDE OIL PRICE UNDER STRUCTURAL BREAKS AND VOLATILITY USING FACEBOOK PROPHET AND HYBRID MODELS

Abstract

Predicting crude oil prices accurately is crucial for effective economic strategies, risk mitigation, and policy development in economies reliant on oil. This research presents an innovative hybrid forecasting system that combines Prophet trend analysis, ARIMA and SARIMA models for modeling volatility, and machine learning techniques for residual learning. This approach aims to thoroughly account for long-term trends, seasonal variations, nonlinear behaviors, volatility clustering, and structural breaks found in Nigeria's daily crude oil prices. We analyze crude daily price spanning from January 1986 to December 17, 2025, to maintain high-frequency data essential for short-term market assessments. Initial time plot indicates significant nonlinearity, heavy-tailed distributions, volatility clustering, and multiple structural breaks connected to major global occurrences such as the oil price crash from 2014 to 2016, the COVID-19 crisis, and ongoing geopolitical conflicts. Tests for stationarity demonstrate that price at levels exhibit non-stationarity, whereas returns series are stationary, supporting the application of ARIMA models. The evaluation of ARIMA and SARIMA models, Prophet (both linear and logistic), machine learning (Random Forest, ANN, MLP), and hybrid methods show that ARIMA and SARIMA models provide better accuracy for short-term forecasts, while Prophet model capture medium- to long-term trends and calendar influences. Hybrid statistical model combinations preserve the predictive capacity of ARIMA, and more complex hybrid systems to enhance forecast reliability in the presence of structural breaks and high level of volatility. When the machine learning techniques are used in isolation, the results show inferior calibration and greater error metrics despite achieving some improvements in managing nonlinear trends. The findings add to the current body of knowledge by illustrating how hybrid frameworks merge statistical precision with adaptable trend and volatility components yield strong and pertinent forecasts for risk-driven crude oil markets. These conclusions are significant for investors, policymakers, and stakeholders in the energy sector who are navigating the challenges of volatile oil markets.