How snivyii/deeptvar-model-for-time-series-data works
DeepTVAR is a Streamlit time-series forecasting application: users prepare data, choose a lag-selection method and model, train or reload a model, then forecast a selected variable and inspect results. The main app excerpt shows preprocessing-related imports, lag search, model definition, save/load, training, and forecast setup, but omits much of the UI and runtime code. The standalone scripts provide separate lag-selection and model-training experiments; README documents only launching the app. Connections among these scripts are not shown, so they are represented as independent alternatives.
Main entry points: DeepTVAR_streamlit_main.py Forecast App
Components 14
Streamlit Forecasting
- Forecast App
DeepTVAR_streamlit_main.py - LSTM Lag Search
DeepTVAR_streamlit_main.py - TVVAR Model
DeepTVAR_streamlit_main.py - Saved Models
DeepTVAR_streamlit_main.py - Forecast Results
DeepTVAR_streamlit_main.py
Lag Selection
- LSTM Lag Search
model/find_best_lag_LSTM.py - Seeded Lag Search
model/find_optimal_lag_use_SEED.py - VAR Lag Search
model/find_best_lag_using VAR.py
Standalone Models
- TVVAR-LSTM
model/TVVAR_1.1_khong_grid_search.py - DeepTVAR MLP
model/DeepTVAR_TH2.py - LSTM Baseline
model/DEEPTVAR_TH3.py - DeepTVAR Likelihood
model/DeepTVAR_TH4_Thinh.py - DeepTVAR Variant
model/DeepTVAR_TH5.py
Other components
- Analyst
Connections 5
- Analyst → Forecast App: uses
README.txt - Forecast App → LSTM Lag Search: searches lags
DeepTVAR_streamlit_main.py - Forecast App → TVVAR Model: trains model
DeepTVAR_streamlit_main.py - Forecast App → Saved Models: saves or loads
DeepTVAR_streamlit_main.py - Forecast App → Forecast Results: configures forecast
DeepTVAR_streamlit_main.py