S&P 500 Stock Price Financial Risk Prediction Using ARIMA, SARIMA and LSTM Models

This study presents a comprehensive analysis of the S&P 500 stock dataset, to explore market trends, sector volatility, and short-term financial risk through price forecasting. Along with Exploratory data analysis (EDA), a short-term (30-day ahead) price forecasting was performed specifically for NVIDIA (NVDA) stock us. Three distinct time series models were used: ARIMA (5,1,0), SARIMA with weekly seasonality, and a deep learning-based LSTM enhanced with a Linear Attention mechanism. The model predictions were compared with each other and evaluated against the actual NVDA price at the end of December 2025. Results indicate that SARIMA provided the most balanced and directionally reasonable forecast, while ARIMA was overly conservative, and LSTM tended to over-extrapolate trends. The analysis highlights challenges in accurate stock price prediction with strong momentum-driven markets.