opendoc-pakistan/pakistan_finance_ml_features

Pakistan Finance ML Features

Separate machine-learning feature tables for Pakistan finance: monthly macro/inflation features, weekly SPI price features, government securities yield features, and interest corridor facility usage features.

Source: OpenDoc Pakistan

Pakistan Finance ML Features

Summary

This dataset contains separate machine-learning feature tables for Pakistan finance. It is not mixed with the RAG corpus. Tables include monthly inflation/macro features, weekly SPI item-city price features, government securities yield features, and interest corridor facility-usage features.

Dataset Details

• —Dataset ID: pakistan-finance-ml-features
• —Version: 0.1.0
• —Category: Economy and Finance
• —Tags: machine-learning, forecasting, features, inflation, spi, bonds, yields, interest-rates, pakistan
• —Language: English
• —Geography: Pakistan
• —File Formats: CSV
• —Records: 35,085 feature rows
• —Maintainer: OpenDoc Data Stewardship Team

Source and Provenance

Features are derived from the separate curated packages:

• —pakistan_inflation_and_prices
• —pakistan_government_securities_auctions
• —pakistan_interest_rate_corridor
• —pakistan_macroeconomic_indicators

No RAG chunks or QA text are included in this package.

License

• —Original License: Mixed official public website sources; explicit open redistribution licenses not found for PBS/SBP, World Bank WDI is CC-BY.
• —Repository License: LicenseRef-PBS-Public-Website, LicenseRef-SBP-Public-Website, and World Bank WDI attribution.
• —Attribution Required: Yes, cite the underlying official sources.

Files

• —monthly_macro_inflation_features.csv: monthly CPI/SPI/WPI features, WDI annual macro joins, lags, rolling averages, and next-month targets.
• —weekly_spi_price_features.csv: weekly item-city average prices with lag, rolling, next-week price, and direction targets.
• —government_securities_yield_features.csv: T-bill/PIB auction yield features with lag and next-auction targets.
• —interest_corridor_daily_features.csv: SBP corridor facility usage features with lag, rolling, and next-observation targets.

ML Notes

• —Features are time ordered.
• —Target columns are explicitly prefixed with target_.
• —Splits are chronological, not random.
• —Lag/rolling features use prior observations only.
• —Blank targets at the end of a group are expected where no future observation exists.

Quality Report

• —Duplicate Primary Keys: 0
• —Blank Primary Keys: 0
• —PII Scan: 0 email/phone regex hits.
• —Known Limitation: Some features inherit source frequency differences: monthly inflation, weekly prices, auction events, and daily/operation corridor records are separate tables by design.

Intended Use

Appropriate for:

• —Inflation direction and level forecasting.
• —SPI item/city price anomaly detection.
• —Auction yield movement modeling.
• —Liquidity facility usage modeling.

Out of scope:

• —Retrieval/citation QA; use pakistan_finance_rag_corpus.
• —Trading or investment advice.

Citation

Pakistan Bureau of Statistics, State Bank of Pakistan, and World Bank WDI sources, curated as "Pakistan Finance ML Features" by OpenDoc Pakistan (OpenDoc), v0.1.0, 2026.