Nine datasets built by going to the source and rebuilding it — regulator portals, provincial price boards, court registers, and concession records. Every figure below is a count of records we hold, not a projection.
Quarterly and monthly condensed financial statements for Indonesian commercial banks, scraped from the OJK portal and rebuilt into a four-layer warehouse: raw observations, bank/variable/time dimensions, a canonical fact table, and regression-ready quarterly panels.
Daily retail food prices at individual market level across all 38 regencies and cities of East Java. Collected market by market rather than as provincial averages, which is what makes within-province dispersion analysis possible.
Farm-gate prices recorded at production centres — shallots in Nganjuk and Probolinggo, broilers in Blitar, chillies in Gending. Paired with the consumer series above, this yields a producer-to-consumer margin measured on the same day in the same province.
National daily price series for strategic food commodities, market by market. The national counterpart to the East Java series — used where the research question needs cross-province comparison rather than within-province detail.
Full-text corpus of BI and OJK regulations across banking, sharia, capital markets, non-bank financial institutions, insurance, and fintech — indexed for search and mapped so that amendments and revocations resolve to the instrument they modify.
Structured case records — parties, claims, procedural chronology, verdict text — harvested court by court from the judiciary’s own case-tracking system. Counts are a snapshot of 11 August 2026 and still rising; coverage is partial, not the full national roster of district courts. Only aggregate patterns are published, and individual names are never surfaced in public output.
Full decision texts from the Supreme Court’s public directory for six of the largest metropolitan district courts, retained as parsed records alongside the raw pages they were built from. Paired with the case register above, this links a docket entry to the reasoning that closed it.
Mining business entities with concession permits, shareholder chains, and sanction history — including foreign ownership links and offshore holding vehicles, reconstructed into an ownership graph rather than a flat list.
Twenty-four months of visitor reviews across tourism destinations, cleaned and canonicalised against OpenStreetMap place records, with a multilingual transformer sentiment layer and a category-level correction for review propensity.
Methodology and replication packages are published openly with each paper. The datasets themselves are not redistributed: most derive from public portals whose own terms govern reuse, and the value HAMPARAN adds sits in the cleaning, reconciliation, and structure rather than in the raw rows. Researchers who want to work with a specific series are welcome to write.
The court corpus covers one district court, not the national system. The producer price series observes production centres, not every regency. Bank coverage before 2019 is thinner than after. These limits are stated in the papers that use them and are not worked around in the analysis.