Comparing ChatGPT and LSTM in predicting changes in quarterly financial metrics

In the financial industry, the ability to predict financial metrics accurately and in a timely manner can significantly impact investment decisions, risk management, and strategic planning. In recent years, machine learning has emerged as a powerful tool for such predictions. This study aims to explore the heretofore underexplored predictive potential of ChatGPT by predicting positive or negative changes in quarterly financial metrics, such as revenue and sales numbers, using textual data from social media. The performance of ChatGPT is compared against Long Short-Term Memory (LSTM) neural network models developed as part of this research. The methodology involves preprocessing large datasets from Twitter concerning major companies such as Amazon, Google, and Tesla, training LSTM models, and prompt engineering for ChatGPT-based predictions. Initial findings indicate that LSTM models can predict quarterly financial metric changes with up to 87% accuracy, significantly outperforming ChatGPT, which achieves a maximum accuracy of 67%. Therefore, at the current time, ChatGPT cannot be considered a reliable predictive tool for changes in quarterly financial metrics.