How wahyuramadhandotcom/a-hybrid-forecasting-model works
This is a research forecasting and evaluation repository: researchers run experiments on pharmaceutical and retail demand data, compare AR-LRX with baselines under a shared time-series protocol, and examine robustness and statistical evidence in saved results. The central workflow separates dataset preparation and evaluation rules from model implementations, then branches into replications, perturbation studies, and inference. README documentation supports the experiment-to-result links; sampled source confirms the shared protocol and AR-LRX implementation. Some experiment notebooks and tool internals were not sampled, so their specific wiring is not inferred.
Main entry points: protocol.py Shared protocol · arlrx.py AR-LRX models · results Saved results · exp12_pharma_baselines_completion.py Pharma baselines
Components 18
Demand data
- PharmaSales
data/raw/pharma-sales - Rossmann
data/raw/rossmann
Forecasting core
- Shared protocol
src/experiments/protocol.py - AR-LRX models
src/experiments/arlrx.py - Baseline models
src/experiments/baselines.py
Experiment studies
- Pharma studies
notebooks/exp05c_pharma_arlrx.ipynb - Rossmann studies
notebooks/exp05b_rossmann_arlrx_audit.ipynb - Baseline comparisons
notebooks/exp06b_rossmann_baselines_strong.ipynb - Rolling replications
tools/exp07_rolling_origin.py - Learner stability
tools/exp08_learner_stability.py - Perturbation tests
tools/exp10_controlled_perturbation.py - Calendar study
tools/exp11_pharma_calendar_s1.py - Pharma baselines
tools/exp12_pharma_baselines_completion.py
Evaluation evidence
- Saved results
results - Dependence inference
tools/groupB_inference.py - Pharma crossfit audit
tools/crossfit_audit_pharma.py - Rossmann crossfit audit
tools/crossfit_audit_rossmann.py
Other components
- Researcher
Connections 34
- Researcher → Pharma studies: runs study
- Researcher → Rossmann studies: runs study
- Researcher → Baseline comparisons: runs study
- Researcher → Rolling replications: runs replication
- Researcher → Learner stability: runs study
- Researcher → Perturbation tests: runs study
- PharmaSales → Shared protocol: provides observations
- Rossmann → Shared protocol: provides observations
- Pharma studies → Shared protocol: uses protocol
- Pharma studies → AR-LRX models: runs models
- Rossmann studies → Shared protocol: uses protocol
- Rossmann studies → AR-LRX models: runs models
- Baseline comparisons → Shared protocol: uses protocol
- Baseline comparisons → Baseline models: runs baselines
- Rolling replications → Shared protocol: uses protocol
- Rolling replications → AR-LRX models: runs models
- Learner stability → Shared protocol: uses protocol
- Learner stability → AR-LRX models: tests learner
- Perturbation tests → Shared protocol: uses protocol
- Perturbation tests → AR-LRX models: runs models
- Calendar study → Shared protocol: uses protocol
- Calendar study → AR-LRX models: runs models
- Pharma baselines → Shared protocol: uses protocol
- Pharma baselines → AR-LRX models: runs model
- Pharma baselines → Baseline models: runs baselines
- Shared protocol → Saved results: writes metrics
- Dependence inference → Saved results: reads predictions
- Dependence inference → Saved results: writes inference
- Pharma crossfit audit → Shared protocol: uses protocol
- Pharma crossfit audit → AR-LRX models: audits residuals
- Pharma crossfit audit → Saved results: writes audit
- Rossmann crossfit audit → Shared protocol: uses protocol
- Rossmann crossfit audit → AR-LRX models: audits residuals
- Rossmann crossfit audit → Saved results: writes audit