Strongly-typed genetic programming with an active trading operator for algorithmic trading using fundamental, sentiment, and technical analysis

Algorithmic trading has advanced significantly in recent years, with investors typically relying on either fundamental analysis (FA) or technical analysis (TA) indicators. Sentiment analysis (SA), however, has only gained substantial attention in the last decade. Consequently, algorithms have become essential tools in creating pre-programmed trading strategies, yielding notable financial gains. Despite the extensive individual study of these three types of analyses, their combined potential remains underexplored. Given the success of FA, TA, and SA in identifying profitable trading strategies, this paper aims to examine whether merging these indicators can further enhance profitability. To this end, we introduce a novel strongly-typed Genetic Programming (GP) algorithm within a soft computing framework, designed to enhance robustness in trading strategy development. Each branch in the algorithm's tree structure is dedicated to one analysis type, meaning that one branch includes only fundamental indicators, another solely sentiment indicators, and the final branch exclusively technical indicators. This structured approach allows for better exploration and exploitation of the indicator search space. Additionally, the algorithm incorporates a unique fitness function, which evaluates not only overall trading performance but also the contributions of the individual FA, TA, and SA branches. Furthermore, we propose an innovative GP operator that promotes active trading by injecting trees into the GP population capable of executing numerous trades while maintaining high profitability and low risk. The algorithm's effectiveness is tested using the stocks of 42 international companies, in which it is compared against eleven alternative algorithms across three key financial metrics: Sharpe ratio, rate of return, and risk. The findings show that the proposed GP algorithm, leveraging evolutionary computing principles, demonstrates practical financial benefits. The algorithm enhances the profitability and risk-adjusted performance of trading strategies, achieving statistically significant improvements over competing models and in active trading potential. This work contributes to the advancement of decision support systems and financial analysis applications.