Quantitative Trading Strategy Research Based On Event Features

There is a large amount of structured data in the financial sector,and with the continuous updates and iterations of machine learning and deep learning algorithms,it has become a trend to use these algorithms in the financial field.Quantitative trading is also becoming more common in A-share markets.In recent years,quantitative private equity funds have emerged one after another,but their trading strategies are not publicly disclosed and often time-consuming.Therefore,quantitative trading is still far away from ordinary individual investors.This paper studies the event characteristics of the stock market and provides a reference for easy-to-implement quantitative investment directions.Based on efficient market theory,this paper uses model algorithms such as TabTransformer and XGBoost to input 428 event features collected manually into the stock market as input variables and predict the stock’s related trend information after an event occurs as output variables.The conclusion is that by analyzing the information of the stock market events,we can predict the stock price trend.By using the SHAP tool,we can also obtain the stock market event features with high importance for each price trend feature,proving that these works are interpretable.Based on this conclusion,this paper constructs a quantitative trading strategy based on event feature information of convertible bonds and macro indices.In backtesting experiments,this strategy can achieve an annualized return of 20%and has some self-correction ability to external policy disturbances.Most importantly,this strategy’s data and models are easy to implement.