Abstract:To improve the identification of abnormal quotation behavior in high-frequency trading scenarios, a multi-timescale abnormal quotation monitoring method is developed based on order-by-order data, trade-by-trade data, and order book data. Following data quality control and time alignment, 18 quotation features are extracted from six dimensions, including order submission intensity, order cancellation behavior, order lifetime, trade conversion, price deviation, and order book imbalance. Abnormal behaviors such as intensive quoting, rapid order cancellation, price deviation, and order book imbalance are then identified. Monitoring results show that the abnormality rate under volatile market conditions is significantly higher than that under stable market conditions, and abnormal events are mainly concentrated within a small number of high-risk time windows. The proposed method can capture changes in abnormal quotation characteristics under different trading conditions, providing data support for real-time screening of abnormal quotations, review of key periods, and risk classification.