Intraday trading strategy based on time series and machine learning for Chinese stock market

In this study by Wang et al., an intraday trading strategy is developed for the Chinese stock market using Markowitz optimization and Multilayer Perceptron (MLP). The researchers select three promising stocks from each of the ten industries, including Science and Technology, Basic Materials, Consumption (periodic), Finance, Service, Means of Production, Energy, Consumption (aperiodic), Health Care, and Transportation. By diversifying across industries, the strategy aims to manage risk and avoid unexpected volatility. The trading strategy involves opening positions for all selected stocks on the first day of trading. Subsequently, trades are executed twice daily to capitalize on intraday volatility during specific trading hours. The continuous bidding hours for A-shares are from 9:30 to 11:30 in the morning and from 13:00 to 15:00 in the afternoon with a break at noon. To predict stock prices for intraday trading, one-minute trading data from 9:30 to 11:20 is used to train an MLP model. This model predicts stock prices at 14:50 based on data up to that point in time. Trading decisions are made based on the price difference between 14:50 and 11:20; if the predicted price is lower than the earlier price, positions are sold before noon and bought back at 14:50. Conversely, if the predicted price is higher than the earlier price, additional positions are bought before noon and sold at 14:50. The study utilizes experimental data including close prices of minute-by-minute data in October as well as daily close prices from January to September of a given year. Daily data is used for Markowitz Optimization while minute-by-minute data is employed for day trading purposes. Overall, this research aims to achieve profitability through intraday trend prediction by combining Markowitz portfolio optimization with MLP-based stock price prediction. The results section includes findings related to portfolio optimization effectiveness, modeling outcomes, and evaluation of overall strategy returns. Future work may focus on improving model performance and comparing it with other existing models in the field.

Intraday Trading Strategy for the Chinese Stock Market: Combining Markowitz Optimization and Multilayer Perceptron

In recent years, with the rise of artificial intelligence and machine learning, there has been a growing interest in using these techniques to develop trading strategies. In this study by Wang et al., the researchers aim to create an intraday trading strategy for the Chinese stock market that combines Markowitz optimization and Multilayer Perceptron (MLP) to achieve profitability.

The Chinese stock market is known for its volatility, making it an attractive market for day traders looking to capitalize on short-term price movements. However, this also presents a challenge as sudden price fluctuations can lead to unexpected losses. To manage risk and avoid such volatility, the researchers select three promising stocks from each of the ten industries in China.

These industries include Science and Technology, Basic Materials, Consumption (periodic), Finance, Service, Means of Production, Energy, Consumption (aperiodic), Health Care, and Transportation. By diversifying across industries, the strategy aims to spread out risk and increase potential returns.

The trading strategy involves opening positions for all selected stocks on the first day of trading. Subsequently, trades are executed twice daily during specific trading hours – 9:30-11:30 am and 1:00-3:00 pm with a break at noon. This takes advantage of intraday volatility while minimizing exposure overnight when unexpected events may occur.

To predict stock prices for intraday trading purposes accurately, one-minute trading data from 9:30 am to 11:20 am is used to train an MLP model. This model predicts stock prices at 14:50 based on data up until that point in time. Trading decisions are then made based on the difference between predicted prices at 14:50 and earlier prices at 11:20.

If the predicted price is lower than the earlier price, positions are sold before noon and bought back at 14:50. On the other hand, if the predicted price is higher than the earlier price, additional positions are bought before noon and sold at 14:50. This approach aims to take advantage of short-term trends while minimizing risk.

The study utilizes experimental data from October as well as daily close prices from January to September of a given year. Daily data is used for Markowitz Optimization, which helps determine an optimal portfolio allocation based on expected returns and risks. Minute-by-minute data is employed for day trading purposes.

The results section includes findings related to portfolio optimization effectiveness, modeling outcomes, and evaluation of overall strategy returns. The researchers found that combining Markowitz optimization with MLP-based stock price prediction resulted in significant improvements in portfolio performance compared to traditional methods.

Future work may focus on improving model performance by incorporating more advanced techniques such as deep learning or comparing this strategy with other existing models in the field. Additionally, expanding this research beyond Chinese stocks could provide valuable insights into applying similar strategies in other markets.

In conclusion, Wang et al.'s study showcases how combining Markowitz optimization with Multilayer Perceptron can create a profitable intraday trading strategy for the Chinese stock market. By diversifying across industries and utilizing minute-by-minute data for day trading decisions, this approach aims to manage risk while capitalizing on short-term trends. With further advancements in technology and machine learning techniques, we can expect to see even more sophisticated trading strategies emerge in the future.