Title: Detect the stock price trend in high-frequency data: A stochastic disorder approach
Authors: Yaosong Zhan - Renmin University of China (China) [presenting]
Abstract: Identifying the trend of a stock price and its changes is critical for the trend following strategy. We propose a model for detecting trend changes in high-frequency data based on the stochastic disorder theory. We use geometric Brownian motion to describe the stock price, and the compound Poisson process is used to solve the jumps that occur in the high-frequency data. Through solving an optimal stopping time problem, we propose a rule for identifying trend changes. The numerical simulation results show that the proposed model detects whether the trend has changed in time with a small error probability and time lag. Compared with the moving-average(MA) technical indicators, the model is more accurate and less likely to identify wrong trend-changing points. We construct a trend-following strategy based on the model, and it outperforms the MA strategy in the Chinese stock market.