Trading Signal Survival Analysis: A Framework for Enhancing Technical Analysis Strategies in Stock Markets

Algorithmic trading is one important financial area of interest to both academic and industrial researchers. With the development of machine learning and deep learning, all kinds of models and techniques are utilized in algorithmic trading. This paper proposes a novel framework for enhancing stock technical analysis strategies by survival analysis. The main idea is to integrate an existing trading strategy with a survival model and make them complementary to each other. By means of survival analysis, the original trading strategy can be extended to introduce an investment target, which is treated as the event of interest. On the other hand, the original trading signal provides survival analysis with a simple and clear starting time point of observation. The trained survival models are used to filter out false trading signals to improve the strategy performance. Under the framework, we propose different filtering methods, utilize different deep survival models, and compare their performance from both trading and model perspectives. We perform extensive and strict backtesting on the daily trading data of 380 plus stocks. The experimental results show that the framework can well improve the performance of technical analysis strategies in different market situations.