High-Performance Machine Learning for Fintech

Abstract

This paper introduces a high-performance compute engine based on differential evolution and tailored for machine learning techniques in the FinTech sector. We demonstrate enhanced runtime performance, which allows for testing a broader array of candidate investment and trading strategies, thereby expanding the scope and improving the quality of strategy evaluations. Serving as the foundation of our differential-evolution-based machine learning framework for portfolio selection and management, this engine is designed for processing real market data and executing highly customisable trading strategies. We present various techniques to optimise its runtime performance, assess their relative impact on performance, and quantify its superior performance compared to existing engines.

Keywords: Algorithmic Trading, Machine Learning, Genetic Programming, StrategyGeneration, Portfolio Selection, Robustness Testing