Smart System for Detecting Anomalies In Crude Oil Prices Using Long Short-Term Memory

Abstract

Crude oil is leading globally, as it represents roughly about 33% of the total energy consumed globally. It is one of the most significant exchanged resources in the world, oil in one way or the other affects our day to day routines, like transportation, cooking and power, and other numerous petrochemical items going from the things we use to the things we wear. The increment sought after for petroleum derivatives is on a persistent ascent, making it vital for the oil and gas industry to think of new methodologies for further developing activity. This paper presents a smart system for detecting anomalies in crude oil prices. The experimental process of the proposed system is of two phases. The first phase has to do with the pre-processing stage, and the training stage while the second phase of the experiment has to do with the building/training of the Long Short-Term Memory algorithm. The experimental result shows that LSTM model had an accuracy result of 98%. The result further shows that our proposed model is under fitting since the training loss is lesser than the validation loss. The proposed model was saved and was used in detecting anomalies of the crude oil prices ranging from 1990 to 2020.