Stock price forecasting has long been considered a challenging field, made even more interesting by the combination of different scientific disciplines. My research presents a quantum-inspired deep machine learning model that integrates conventional machine learning methods with quantum circuits to improve forecasting accuracy.

The aim of the research is to process historical trading data, sentiment-based news, and market indicators from different data sources within a hybrid model. The architecture of the proposed model incorporates a transformer neural network enhanced with quantum-inspired layers, enabling the efficient identification of complex patterns.

The implementation uses Python and the TensorFlow framework for the deep learning model, while IBM Qiskit is used to implement the quantum circuits. The model is trained on historical data obtained from the CRSP US and Yahoo! Finance databases, as well as news articles processed through sentiment analysis and automatically collected using the NewsAPI. The results show that the transformer model enhanced with a quantum-inspired approach improves forecasting accuracy compared to existing conventional deep learning models. The uniqueness of the model enables it to achieve highly accurate predictions even when using smaller datasets, forecasting stock closing prices with an accuracy exceeding 1 percentage point compared to the baseline models, thus offering significant potential advantages in stock market forecasting.

Another notable aspect of the research is its scalable architecture, which operates through Microsoft Azure cloud services, ensuring broad applicability of the solution. This approach provides a new perspective on financial data processing and predictive modeling.