Detecting Potential Market Corner Risk of WTI: A Hybrid Algorithm Anomaly Detection Approach

Abstract The price of crude oil is one of the most critical factors of the world economy, as it is volatile and sensibly affected by the macro-economic, thus attracting large-scale speculative activities. Well known for its vulnerability, some price fluctuations are affected by external factors, such as the financial crisis and oil trade war, while others are due to manipulation. Due to such facts, this article aims to clarify the characteristics and features of the market corner in the crude oil future market and detect potential market corner risk in West Texas Intermediate (WTI), and a hybrid anomaly detection approach is proposed. After a detailed overview of the characteristics and definitions of the market corner, the features are extracted by processing actual market data. We first detect suspicious market corners with abnormal prices and volume through the Local Outlier Factor algorithm (LOF). Next, the detected results are used as pseudo-labels, and the entire month's trading behavior is trained and classified through the Support Vector Machine (SVM) and Random Forest (RF) algorithms to identify potential market corners.Experimental results show that the proposed model has excellent accuracy, precision, recall, and F-score, indicating that the model is feasible and has strong robustness. Furthermore, based on the successful detection of potential market corner risk, the model can be further used for individual risk control and overall supervision of the crude oil futures market.