Makassar, Indonesia, August 25th, 2025/asds/ Financial markets are often described as unpredictable, fast-moving, and complex. Stock prices rise and fall as a result of countless interacting factors, from company performance and investor sentiment to global economic shocks. For decades, analysts have searched for reliable ways to understand these movements. In recent years, artificial intelligence (AI)—especially deep learning—has emerged as one of the most promising tools to help make sense of market dynamics. A notable contribution to this field is a recent study published in IEEE Access that explores how advanced neural networks can significantly improve stock price forecasting accuracy.
At the heart of the study is a simple but powerful idea: instead of relying on a single AI model, why not combine the strengths of several? The researchers introduced a compositional learning framework, which allows different types of Recurrent Neural Networks (RNNs)—namely Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Simple Recurrent Unit (SRU)—to work together. Each model processes financial time-series data in slightly different ways, making the combined system more flexible and robust when dealing with real-world market volatility.
What makes this approach particularly innovative is how the data are handled. Rather than feeding all stock indicators into the model at once, the system first learns patterns from each variable separately—such as opening price, closing price, and trading volume—before merging them into a unified prediction. This design helps the model capture subtle trends that are often missed by traditional forecasting methods. To further enhance performance, the researchers employed an optimization technique inspired by nature, known as the Grey Wolf Optimizer, which fine-tunes the model’s parameters in a highly efficient way.
The results are impressive. When tested on long-term stock market data, the optimized LSTM-based model achieved extremely high prediction accuracy with very low error rates. In practical terms, this means predictions that are closer to real market movements and less prone to sudden miscalculations. Such accuracy is not only valuable for investors, but also for institutions managing financial risks, pensions, or development programs that depend on stable economic planning.
Beyond finance, the implications of this research connect directly to the Sustainable Development Goals (SDGs). By improving data-driven decision-making, this work supports SDG 1 (No Poverty), as better financial forecasting can contribute to more stable economic growth and informed investment strategies that benefit society at large. It also aligns strongly with SDG 4 (Quality Education) by advancing knowledge in data science, artificial intelligence, and applied mathematics—fields that are increasingly essential for future professionals. Most importantly, the study reflects the spirit of SDG 17 (Partnerships for the Goals), as it demonstrates how collaboration between mathematics, statistics, and computer science can produce impactful solutions to complex global challenges.
In a world where uncertainty is the norm, smarter forecasting tools offer a way forward. This research shows that when advanced algorithms are carefully designed and responsibly applied, AI can become a powerful ally in building more resilient, informed, and sustainable economic systems.
DOI: 10.1109/ACCESS.2025.3602721
Contact: Prof. Armin Lawi
+6281355794212
armin@unhas.ac.id




