تشخیص دستکاری بازار بر اساس روش های طبقه بندی

In this paper, we use supervised machine learning methods to detect the market manipulation in China based on the information released by China Securities Regulation Commission (CSRC) and data in the security market. Among the supervised machine learning, we mainly use classification methods to detect the anomaly from the daily and tick trading data of manipulated stocks. As a result, we find that the supervised machine learning methods are good at detecting market manipulation from daily trading data and have poor performance on tick data, based the measure method of accuracy, sensitivity, specificity and area under the curve (AUC). The best used supervised machine learning models are K-Nearest Neighbor (KNN) and Decision Tree (DT) which have over 99% of them.