Machine LearningFull-Stack
Eight forecasting models behind one Streamlit interface.
Streamlit app and experiment notebook. Requires a user-supplied price CSV.
StreamlitstatsmodelsProphetpmdarimaTensorFlowPlotly
Overview
Upload a price series and explore SMA, EMA, RSI, MACD, decomposition, and stationarity tests, then compare ARIMA, SARIMAX, ETS, Prophet, SVR, hybrid ARIMA-ANN, ANN, and LSTM forecasts.
Recorded results
- Rolling ARIMA(3,1,3) test MSE
- 2,195.88
- S&P 500 closes, split at 2019
- Source: maincode.ipynb, cell 70
Limitations
- Some neural baselines in the notebook use test data during training or prediction, so their errors are optimistic.