How karlheinrich-jpg/the-data-of-energy-poverty-index works
This repository is a research workflow for constructing a provincial energy-poverty index, forecasting it from province-year predictors, interpreting model predictions, and comparing those patterns with fixed-effects panel associations. The documented main path runs from indicator data through entropy-weighted EPI and chronological SSA-LSTM evaluation; benchmark models and alternate index constructions support robustness. Source excerpts are partial, so links to unsampled analyses are based on README descriptions and are limited to documented relationships.
Main entry points: Exogenous_Variables_and_Table.xlsx Predictor data · EPI_entropy_output.xlsx EPI workbooks · LSTM-SHAP-base.py LSTM SHAP · SSA-LSTM.py SSA-LSTM
Components 19
Index construction
- Indicator data
EP数据.xlsx - Entropy EPI
能源贫困预测CNN-LSTM/EWM.py - EPI workbooks
能源贫困预测CNN-LSTM/EPI_entropy_output.xlsx - Alternate EPI
能源贫困预测CNN-LSTM/CRITIC-因子分析-PCA.py - Alternate index results
能源贫困预测CNN-LSTM/EPI_CRITIC_FA_PCA_output.xlsx
Forecasting
- Predictor data
能源贫困预测CNN-LSTM/Exogenous_Variables_and_Table.xlsx - SSA-LSTM
能源贫困预测CNN-LSTM/SSA-LSTM.py - Model checkpoint
能源贫困预测CNN-LSTM/best_SSA_LSTM_checkpoint.pth - Model benchmarks
能源贫困预测CNN-LSTM/模型对比.py - Comparison plots
能源贫困预测CNN-LSTM/模型对比＋绘图.py - Benchmark results
能源贫困预测CNN-LSTM/Model_Comparison.xlsx
Interpretation
- LSTM SHAP
能源贫困预测CNN-LSTM/LSTM-SHAP-base.py - Regional SHAP
能源贫困预测CNN-LSTM/LSTM-SHAP-异质性.py - Index robustness SHAP
能源贫困预测CNN-LSTM/LSTM-SHAP-鲁棒性.py - SHAP evidence
能源贫困预测CNN-LSTM/SHAP分析+TOP依赖图.py
Panel inference
- TWFE analysis
能源贫困预测CNN-LSTM - TWFE results
能源贫困预测CNN-LSTM/TWFE_EPI_results.xlsx
Other components
- Researcher
- Research results
Connections 35
- Researcher → Entropy EPI: constructs index
- Indicator data → Entropy EPI: provides indicators
- Entropy EPI → EPI workbooks: writes EPI
- Researcher → Alternate EPI: tests alternatives
- Indicator data → Alternate EPI: provides indicators
- Alternate EPI → Alternate index results: writes scores
- Researcher → SSA-LSTM: runs forecast
- EPI workbooks → SSA-LSTM: supplies target
- Predictor data → SSA-LSTM: supplies predictors
- SSA-LSTM → Model checkpoint: saves weights
- Researcher → Model benchmarks: runs comparisons
- EPI workbooks → Model benchmarks: supplies target
- Predictor data → Model benchmarks: supplies predictors
- Model benchmarks → Benchmark results: writes metrics
- Benchmark results → Comparison plots: provides metrics
- Researcher → LSTM SHAP: interprets forecasts
- Model checkpoint → LSTM SHAP: provides model
- EPI workbooks → LSTM SHAP: supplies target
- Predictor data → LSTM SHAP: supplies features
- Researcher → Regional SHAP: tests regions
- EPI workbooks → Regional SHAP: supplies target
- Predictor data → Regional SHAP: supplies features
- Researcher → Index robustness SHAP: tests robustness
- Alternate index results → Index robustness SHAP: supplies alternate targets
- Predictor data → Index robustness SHAP: supplies features
- LSTM SHAP → SHAP evidence: supports interpretation
- Regional SHAP → SHAP evidence: supports interpretation
- Index robustness SHAP → SHAP evidence: supports interpretation
- Researcher → TWFE analysis: runs panel analysis
- EPI workbooks → TWFE analysis: supplies outcome
- Predictor data → TWFE analysis: supplies lagged predictors
- TWFE analysis → TWFE results: writes estimates
- SHAP evidence → Research results: contributes evidence
- TWFE results → Research results: contributes estimates
- Benchmark results → Research results: contributes metrics