Developing a Probabilistic Modeling Algorithm for Movement Patterns in Financial Markets Using Data Mining and Optimal Data Size Determination (A Comparative Study with ARIMA-(G)ARCH econometric models)

Abstract

In the volatile realm of financial markets, this research outlines a roadmap to enhance the accuracy of financial market trend predictions by proposing an algorithm based on the conditional probability distribution of sequential movement patterns and data mining techniques to determine the optimal size of probabilistic modeling data. The study demonstrates the algorithm’s efficiency and generalizability through its application to multiple global and Arab financial markets (14 financial indices from global and Arab stock markets).

The key findings of this research highlight the superiority of the proposed algorithm—which incorporates optimal data size determination—over both the algorithm without optimal data size determination and ARIMA-(G)ARCH econometric models. This conclusion is supported by the application of the Predictive Model Value-Added criterion, which provides an intrinsic evaluation of the utility of any applied predictive model. Additionally, a statistically significant difference analysis was conducted between: (1) the average prediction accuracy of the proposed algorithm and econometric models, and (2) the proposed algorithm with and without optimal data size determination. All applied comparisons and criteria conclusively demonstrated the superiority of the proposed algorithm with optimal data size determination, confirming its stability, consistency, and generalizability. These innovations open new avenues for refining investment strategies and mitigating risks, establishing the algorithm as a valuable tool for investors and financial analysts.