Use this URL to cite or link to this record in EThOS:
Title: Machine learning based trading strategies for the Chinese Stock Market
Author: Du, Juan
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2020
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis focuses on the machine learning based trading strategies of China Exchange Traded Funds (ETFs). Machine learning and artificial intelligence (AI) provide an innovative level of service for financial forecasting, customer service and data security. Through the development of automated investment advisors powered by machine learning technology, financial institutions such as JPMorgan, the Bank of America and Morgan Stanley have recently achieved AI investment forecasting. This thesis intends to provide original insights into machine learning based trading strategies by producing trading signals based on forecasts of stock price movements. Theories and models associated with algorithmic trading, price forecasting and trading signal generation are considered; in particular machine learning models such as logistic regression, support vector machine, neural network and ensemble learning methods. Each potential profitable strategy of the China ETFs is tested, and the risk-adjusted returns for corresponding strategies are analysed in detail. The primary aim of this thesis is to develop two machine learning based trading strategies, in which machine learning models are utilised to predict trading signals. Each machine learning model and their combinations are employed to generate trading signals according to one day ahead forecasts, demonstrating that the final excess return does not cover the transaction costs. This encourages us to reduce the number of unprofitable trades in the trading system by adopting the 'multi-day forecasts' in place of the 'one day ahead forecasts'. Therefore, investors benefit from a longer prediction horizon, in which more predicted information of the total number of upward (or downward) price movements is provided. Investors can make trading decisions based on the majority of the predicted trading signals within the prediction horizon. Moreover, this method of trading rules is consistent with the industry practice. The strategy is flexible to allow risk-averse investors and risk-loving investors to make different trading decisions. A multi-day forecast based trading system through random forest yields positive risk-adjusted returns after transaction costs. It is identified that it is possible that some machine learning techniques can successfully assist individuals in navigating their decision-making activities.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral