SPY(ETF) Price Forecaster

Workflow

• Fetched SPY market data and engineered MA10/30/90, log returns, volatility, and RSI14 features.
• Built sequence windows (90, 30, and 10 days) so each sample gives the model recent price/indicator history.
• Each LSTM learns temporal patterns by carrying memory from earlier timesteps and outputs a next-day price forecast.
• Compared all models using MSE, RMSE, and MAE, then benchmarked them against a MA10 baseline.
• Generated next-day predictions and plotted model outputs against actual prices.
• In simple terms: the model looks at recent market behavior, learns recurring patterns, and makes an informed guess for tomorrow's price.