POTENTIAL STOCK PRICE TREND PREDICTION USING GENERATIVE AI MODEL

The rapid advancement of generative AI offers notable implications for investment decision-making, yet studies utilizing financial ratios to predict stock prices remain limited. This research aims to evaluate the potential of AI models ChatGPT, Gemini, Deepseek, and Claude in forecasting LQ45 stock price trends using financial ratios and historical data, while also testing the consistency of their predictions over time. Employing an experimental quantitative approach, this study analyzes predictions made by four AI models for 23 LQ45-listed companies during the 2021–2023 period. Robustness was assessed by administering identical prompts at two different times and analyzing the results using the Paired Sample t-Test. Accuracy was evaluated at two levels: trend prediction accuracy (Level 1) and price prediction error (Level 2). The findings reveal that while AI models show relatively stable performance in trend direction prediction, their accuracy varies across models. Forecasting exact stock prices remains challenging, indicating AI's current limitations as a fully reliable predictive tool.