CHAPTER ONE

INTRODUCTION

1.1 BACKGROUND OF THE STUDY

The global economy relies on crude oil due to its significance as a production factor and its emergence as the focal point of numerous industrial activities. Oil price fluctuations pose significant influences on the real economy and therefore are interconnected with the various constituents of the global economy. The stock markets and Foreign exchange rates are the main markets which are inclined to very much associate with ever-changing oil prices (Abdullahi & Rui, 2020). In the present post-covid19 times, the price of crude oil has presented pronounced volatility; the surge in price has obviously spawned ample research interest especially in the dynamic interrelatedness existing amongst the price of oil, foreign exchange rates and the stock markets. According to Abdullahi & Rui (2020), this relationship is aforethought to be precarious for variations surrounding global business cycles, economic activity inconstancies, trade gap and remarkable irregularities in financial markets. Therefore, the price of crude oil is always a diligently examined indicator for policy makers and investors alike to enforce fitting risk management strategies and policy design.

On the developing African viewpoint, Pershin, Molero& de Gracia (2016)reveal that the deportment of the foreign exchange rates in several African countries are contrasted; and volatilities of oil price may not primarily be mirrored in foreign exchange rate markets. Korley&Giouvris (2021) revealed some solid proof of the interconnection between domestic stock prices and foreign exchange rates in Nigeria with no causality between stock prices and exchange rates in South Africa, being the largest economies in Sub-Saharan Africa.  For the largest exporter of crude oil (Nigeria), Babatunde (2015) exposed positive but insignificant response of the Nigerian stock market returns to oil price volatilities relapses to negative effects and volatility of oil prices depresses the stock market returns. This signifies that oil price volatilities are indeed an important source of volatility in the Nigerian stock market. Salisu& Mobolaji (2013) also expressed bidirectional returns and volatility transmissions between the foreign exchange markets in Nigeria and oil price.

The annual budget of Nigeria has continuously been clinched to a specific sum of the international crude oil price. Hence, making both the government fiscal and monetary policy to be prone to instability that may arise due to volatilities associated with crude oil price; which stimulates the performance of the economy through the foreign exchange rate (Olaniran, 2019). Forecasting the price of crude oil is intricate due to the high uncertainties associated with price of crude oil. Crude oil price is greatly irregular, non-linear and fluctuates vigorously (Peng, Li & Drakeford, 2020). The peculiar nature of crude oil price i.e. its instability is largely tied to a number of factors such as demand and supply in the financial market, economic growth as well as the technological advances that birth a complex relationship between these factors and the price of oil (Zhao, Jianping& Lean 2017). There exists un-ending deliberation among economists and researchers over the influence of crude oil price instabilities on the performance of the global economy. This fuels major concern to research communities, institutions and government agencies alike. Hence, there exists the need to forecast crude oil price as accurately as is possible. Forecasting prices of crude oil accurately fascinates economists and academic researchers these days, due to crude oil being a source of energy worldwide; and its price fluctuations can influence aggregate economic activities (Pak, 2017).

1.2 STATEMENT OF THE PROBLEM

For decades, forecasting the price of crude oil has lured researchers in the domain of AI (Kaya, 2021). Despite the existence of research related to crude oil price forecasting, there is still a paucity of models for forecasting future crude oil prices in all circumstances (Yin, Peng & Tang, 2018). Multiple approaches for forecasting the price of crude oil using machine learning algorithms have been projected in the past (Li, Shang & Wang, 2019; Zhang &Hamori, 2020). However, they have not been able to solve common drawbacks such as time-wasting, measured convergence, local minima and overfitting(Li, Hu, Heng & Chen, 2021). Attaining a crude oil price forecast with an accuracy level of 100% is almost impossible. For that reason, there exists the need to advance the precision of nonlinear and non-stationary time series data sets using predictive analytics approaches (Nonejad, 2020).

Additionally, forecasting crude oil price is affected by external factors like the supply-demand gap of the crude oil itself. The Organization of the Petroleum Exporting Countries (OPEC) policy of production management, supply disruption, and recovery from the global pandemic in recent times could all pose unforeseen demand or refusal for crude oil, consequently upsetting its price (Fazelabdolabadi, 2019). Therefore, an efficient forecasting system to analyze near-future crude oil prices would prove immensely valuable to bridge the demand gap.

1.3 RESEARCH QUESTIONS

This research aims to answer the following questions:

1. What analytical tools can be deployed to forecast the price of crude oil in the Nigerian marketplace?
2. What are the criteria used to evaluate the performance of the proposed model over prevailing models?
3. What other criteria (which were not captured in prevailing work) can affect the performance of the proposed model?

1.4 AIM AND OBJECTIVES

This research aims to propose an ensemble forecast model using complex network analysis and deep learning. This is in an attempt to boost the performance of current crude oil price forecasting in Nigeria, as regards precision, the total time spent as well as overall reliability and to eventually carry out a comparative analysis on the proposed model’s performance against prior models.

The main objectives of the research are to:

1. Use complex network analysis to eliminate noise and nonlinearity in the chosen dataset during preprocessing.
2. Use a deep learning algorithm to develop an ensemble model for forecasting crude oil prices.
3. Assess the efficiency of the proposed model.

1.5 SIGNIFICANCE OF THE STUDY

It is difficult to ascertain the impact of research that is yet to be carried out. One can only merely anticipate. It is anticipated that this research could go a long way in making significant scholarly contributions to both theories and practice most especially in the field of supervised machine learning algorithms as well as the oil and gas sector in Nigeria. Government agencies, research institutions and even policymakers are in the cards of applying the proposed approach to largely inform their decision making (Valdman & Malyarenko, 2020). Lastly, it is anticipated that the proposed model shall attain an accuracy level that beats current crude oil price forecasts in Nigeria, thereby proving its reliability.

1.6 SCOPE AND LIMITATION

The limitation of the proposed work is to formulate an ensemble forecasting model for the crude oil prices in Nigeria using deep learning and complex network analysis. This work shall not only consider crude oil price in Nigeria but at least one other factor with the potential to pose effects on the price of crude oil such as the dollar exchange rate, demand and supply etc. Past works have been majorly based on monthly data, which restrict the prediction horizons to months. For improved accuracy levels, the proposed model shall be based on weekly historical data. The proposed technique may also be applied in other fields that make use of time series data which are usually irregular, nonlinear and complex for instance the stock market data for accurate forecasting.

1.7 JUSTIFICATION OF THE STUDY

This research is largely motivated by Nigeria’s heavy reliance on the Oil Sector as an economy.  Observations on certain economic factors in Nigeria highlights the following challenges as well as the requirement for an efficient prediction technique:

1. Instabilities associated with the price of crude oil intensely influence Nigeria’s economic development daily.
2. Related studies have exposed the nonlinear nature of economic and financial data, where conventional approaches such as linear prediction models failed to analyze the intricate nonlinear dynamics involved.
3. There is a need for a real-time update of recommended models especially across multiple factors that impact the pricing of crude oil to be generated.

This research is additionally motivated by the likelihood of improving the accuracy and efficiency of crude oil price forecasting through distinguishing the weaknesses in the current algorithms and improving on them (Aamir &Shabri, 2018). Using an ensemble model to forecast crude oil price could help diminish some of the uncertainties known to be linked to budgetary purpose, economic planning of fiscal as well monetary policies of the government of the Federal Republic of Nigeria (Patrick, 2019).

1.8 DEFINITION OF TERMS

Abbreviation                                   Description

AI                                             Artificial Intelligence

ANN                                         Artificial Neural Network

DNN                                         Deep Neural Network

GDP                                         Gross Domestic Product

SSL                                          Semi Supervised Learning

LSTM                                               Long Short-Term Memory

ML                                           Machine Learning

ARIMA                                             Auto Regressive Integrated Moving Average

WTI                                          West Texas Intermediate

WNN                                        Wavelength Neural Network

RNN                                         Recurrent Neural Network

RW                                           Random Walk

SVM                                         Support Vector Machine

EMD                                                Empirical Mode Decomposition

EEMD                                              Ensemble Empirical Mode Decomposition

BPNN                                               Back Propagation Neural Network

OPEC                                               Organisation of the Petroleum Exporting Countries

CBN                                         Central Bank of Nigeria           

1.9 ORGANIZATION OF THE THESIS

This thesis is organized into five (5) chapters. A summary of the concepts of these chapters follows:

Chapter 1 includes a background on the proposed topic, the research questions, the aim and objectives, the significance of the study as well as the scope and limitation of the proposed research.

Chapter 2 provides a literature review with regards to techniques for crude oil price forecasting, analyses of multiple algorithms, existing methodologies and approaches for forecasting crude oil price. This chapter highlights just how literature contributes to the research area and the prevailing approaches that have been used.

Chapter 3 explains the methodology that is to be adopted in the development of the proposed model. This includes the source of historical data used to be used for training, a theoretical overview of the methodology as well as the evaluation processes that shall be applied for comparative analyses.

Chapter 4 covers the implementation of the proposed prediction technique in python programming environment, the results, the comparative analyses as well as discussions of the given results.

Chapter 5 provides the final summary of prior chapters, conclusion, and future recommendations for improvements that can be achieved by means of the results and findings of this thesis.