469 monthly observations of Brent crude oil priced in Indian rupees from May 1987 to May 2026, constructed from public FRED series. The deposit contains replication data and results for a working paper comparing statistical and machine-learning forecast models, including forecast-accuracy metrics, Diebold-Mariano test results, and XGBoost feature-importance output. It was authored by Adesh Mishra and released on Harvard Dataverse in 2026 under a CC BY 4.0 license.
Use Cases
- Benchmarking forecast accuracy of statistical vs. machine-learning models based on the described monthly Brent-INR series.
- Conducting Diebold-Mariano statistical tests for forecast comparison using the included test results.
- Analyzing feature importance for XGBoost models in a financial forecasting context based on the provided output.
- Replicating the econometric analysis from the associated working paper using the provided panel data and results.
Strengths
- 469 months of consistent time-series data spanning from May 1987 to May 2026.
- Includes specific forecast-accuracy metrics and statistical test results (Diebold-Mariano) that support replication.
- Source data is public and re-fetchable from FRED, enhancing transparency.
- Released under a permissive CC BY 4.0 license.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count for the analysis panel is known, but the number of rows in supporting results files is unknown.
- Description metadata is limited; actual data structure and completeness require manual inspection after download.
Provenance
- Source
- Constructed from public Federal Reserve Economic Data (FRED) series: DCOILBRENTEU, DCOILWTICO, DEXINUS.
- Collection Method
- Data was compiled and used for a walk-forward backtest methodology described in the accompanying README.
- Time Range
- May 1987 to May 2026
- Freshness
- Last updated 2026-06-02 18:34:53.