Designing Robust Trading Strategies for High-Frequency Markets Using Big Data and Advanced Computational Techniques

Financial positions are held very briefly in high-frequency trading (HFT) markets and transactions therefore take place very quickly. Success in these markets greatly depends on the effectiveness of trading strategies to maximise returns and minimise risk. Accordingly, this research explores the performance of different machine learning models on simulated datasets that simulate practical market conditions and develop trading strategies that are robust to HFT markets. In the context of quantitative simulation, based on high-frequency trading data, this research is conducted using Python in Google Colab. Advanced machine learning models such as Deep Learning Neural Networks (DLNN), Support Vector Machines (SVM) and Random Forest (RF) are implemented and evaluated using accuracy, precision and recall. With regards to classification in high-dimensional trading data, the SVM model was superior with a precision of 0.53 and recall of 0.32, implying balance. These models, Random Forest & DLNN have shown to have a moderate performance, with potential areas to optimise. Such findings indicate that SVM is a viable technique for predictive trading strategies for high-frequency trading data. It would be interesting to develop future studies that limit the predictive strengths of various machine learning techniques in concert with the prediction quality of both CAOs and soft logic. The findings of this study can immediately have an impact on the future development of algorithmic trading strategies and decision-making and risk management in financial markets. As the data used in this research is synthetic, it is a primary limitation of this research as market conditions may not be captured well.