Article Preview
TopIntroduction
Aided by the scientific progression as well as continuing growth of worldwide global financial incorporation, the demand and supply involving crude oil are determined by escalating sophisticated and various economic contributors globally. This situation, combined with a lot of significant components, which include climate, governmental consistency, economic potentials, consumer anticipations, along with business indicators, appears to have resulted in the varying price activity in the crude oil market. It thus tends to have increased recently, as demonstrated by the activities in benchmark indices, such as the West Texas Intermediate (WTI) and Brent crude oil prices, in association with the trend connected with liberalization and also globalization that happens to be beyond the informative expertise offered by the existing models and techniques. At the same time, because crude oil is actually traded less often as compared to equities, this led to greater degrees of market flaws and comparatively lesser degrees of effectiveness. This results in theoretically important and demanding research issues (Zhu et al., 2014). There was stability in the market prices of crude oil as at early 2000, however, fluctuation changes in the market began to intensify and accelerate afterwards and has not subsided ever since (Askari & Krichene, 2008; He et al., 2016).
Due to these fluctuations in the price of crude oil, experts from the industry and academics have engineered and constructed models as well as techniques in accurately predicting and forecasting the crude oil prices as a way of regulating its movements. The engineering of evolutionary and data mining techniques provided the flexibility of diffusing problems arising with crude oil price. Products of such evolutionary techniques are algorithms that mimic the Biological Immune System (BIS) known as the Artificial Immune System (AIS). Assessment of AIS algorithms are expatiated in Dasgupta et al. (2011). Several other evolutionary algorithms are particle swarm optimization (PSO) (Garg, 2014a; Garg et al., 2014), artificial bee colony (ABC) (Shah et al., 2012; Shah et al., 2014; Garg, 2014b), cuckoo search (CS) algorithm (Garg, 2015b), and biogeography-based optimization (BBO) algorithm (Garg, 2015c). With respect to the aforementioned algorithms, a hybrid gravitational search algorithm (GSA) together with genetic algorithm (GA) was developed for both unconstraint optimization and industrial uncertain data usage (Garg, 2015a; 2019). For the prediction of stock market prices, Shah et al. (2018) came up with a quick gbest guided artificial bee colony as a training process for feed forward neural network termed (QGGABC-FFNN).
The issues arising from the above instability generated from prices of crude oil motivated this research. This study thus proposes the use of Negative Selection Algorithm (NSA) specifically the Variable-Sized Detectors (V-Detectors) (Ji & Dasgupta, 2004; Lasisi et al., 2016), coupled with Fuzzy Rough Feature Selection (FRFS) (Jensen & Shen, 2009), in predicting crude oil price. The purpose of the Fuzzy Rough Feature Selection (FRFS) is to select the best attributes from the crude oil dataset. The reason for selecting the best features is for identifying the most useful and relevant attributes to predicting crude oil price. The selected features from FRFS is consequently trained and tested with the V-Detectors algorithm which aid the performance of the prediction outcome. The objective of this study is to trigger a boost in the performance of V-Detectors via the use of FRFS in selecting best features that represents the overall crude oil price dataset.