API Documentation

Available metrics: call GET /v1/rankings for the full list

Rent estimation using a LightGBM + XGBoost ensemble model (v5), trained on 385K rental listings nationwide with 13 integrated data sources (97 features). Urban MAPE = 2.1%, rural MAPE = 2.6%.

Parameters (JSON body)

Omitted optional fields auto-fill with the same defaults used by the web UI. Applied defaults are returned in input_summary.defaults_applied.

Example Request

Example Response

Model Accuracy

See the accuracy validation page for details.

LightGBM model (phase3_v1) trained on 22 features: macro indicators (CPI, policy rate, housing starts, construction cost, job-openings ratio, FX) plus future population estimates and regional characteristics. Land price change is used as a proxy for rent change (correlation 0.7-0.8 per Japan Real Estate Institute data).

Parameters (JSON body)

Request example

Response example

Model accuracy

Note: not investment advice. Use as a starting point for market sensitivity analysis. Direct rent time-series is not available, so land-price change is used as a proxy.

Combines Phase 2 rent estimation (v5 ensemble, MAPE 2.1%) with Phase 3 land price change forecast to return rent 1 or 3 years out. Useful as a starting point for medium-term yield simulations and investment screens.

Parameters (JSON body)

Request example

Response example

Cross-sectional search over properties owned by all 73 J-REITs. Returns 80+ fields per property: acquisition price, appraisal value, cap rate, occupancy, rental income, land/floor area, structure, ownership type, acquisition-time appraisal, seller (previous owner), and more.

Parameters

Key Response Fields

Example Request

Field coverage varies by property. See the property modal on the web UI for per-field fill rates.

Property disposition events extracted from annual securities reports. Includes sale date, sale price, appraisal at sale, book value, gain/loss, and buyer.

Parameters

Example Request

Hotel-property ADR (average daily rate), RevPAR, GOP, and room occupancy, extracted from annual securities reports. Monthly time-series with currency-split (JPY/USD). Covers Invincible, Hoshino Resorts REIT, JHR, and other hotel REITs.

Parameters

Example Request

Detailed period-level appraisals for J-REIT properties: direct cap price and cap rate, DCF price, discount rate, terminal cap rate, and appraisal NOI.

Parameters

Appraiser-level cap rate bias aggregated as deviation from peer median within use_type × region × year. Positive = higher than peer median (conservative), negative = lower (aggressive). Aggregation dimensions: overall / use_type / region / use_type_region.

Parameters

Time-series detection of appraisal value declines and cap rate spikes for J-REIT properties. alert_level: severe (-10%+ drop), alert (-5%+ drop), watch (-2%+ drop), cap_rate_alert (+0.3pp cap rate), combined_alert (combined signals), suspicious_data (-40%+ drop, likely partial sale or unit mismatch). Intended for activist-investor screening and impairment candidate discovery.

Parameters

Spatial join of 250m-mesh future population projection (XKT013, 2020-2050, based on Japan Census) with REIT property addresses via GSI geocoding. Quantifies population-decline impact on REIT portfolios. weighted_pop_change_2050 is acquisition-price-weighted 2025→2050 population change. pct_aum_high_risk is AUM-weighted share of high-risk properties.

Parameters

risk_level classification (on property detail)

• severe: population drops more than -30% by 2050
• high: -30% to -20%
• medium: -20% to -10%
• low: -10% to 0%
• growth: flat or increasing

Returns latest-period geocoded coordinates and 250m-mesh population projections (2025/2030/2040/2050) for a specific property. Example: /v1/reit/property/OtemachiFirstSquare/demographic-risk

Top tenant information per J-REIT property: tenant name, industry, leased area, contract form, remaining years. Includes anonymized tenants.

Parameters

J-REIT property acquisition events served in a unified transaction-case format. Includes acquisition date, price, price per tsubo (by floor area and by land area), seller, appraisal at acquisition, acquisition-time NOI cap rate, sponsor-transaction flag, already-sold flag, and holding years.

Parameters

Example Request

Master table of J-REIT sponsor companies. Fields include sponsor company name, EDINET code, and match keywords. Used as the reference for sponsor-transaction detection.

Parameters

Normalized master of sellers, buyers, and SPCs that appear in J-REIT transactions. Includes transaction count, related REITs, aliases, and resolved corporate numbers (via the EDINET link, when available). Useful for identifying major buyers/sellers and deduplicating SPCs.

Parameters

Annual timeline joining each J-REIT sponsor's EDINET consolidated financials (revenue, operating income, net income, ROE, YoY deltas) with related-REIT acquisition and disposition events. Useful for detecting patterns like "properties concentrated to the sponsor during a decline year" (sponsor_in_decline=true) — the kind of read an activist analyst would want to verify.

Parameters

Example Request

Predicts the likelihood of a sponsor transferring properties to its affiliated J-REITs in the next fiscal year. Calibrated Logistic Regression (v1) trained on the sponsor performance × REIT acquisition timeline. Leave-One-Sponsor-Out evaluation: AUC=0.75 / Brier=0.05 / ECE=0.009. Features: (1) prior-3y sponsor-to-REIT transfer count (2) sponsor net income YoY (3) revenue YoY (4) years since last transfer (5) prior-3y REIT acquisition activity. Returns per-feature contribution (log-odds) and confidence label (high/medium/low). Treat as a "signal score", not a literal probability (only 13 positive training examples). Design doc: docs/ml_sponsor_pipeline_design.md.

Parameters

Example Request

Municipality-level hazard zone aggregation around J-REIT property locations. BOUSAI DB (area_hazard_risk, 129,102 records) is joined to REIT properties on municipality code. Returns steep-slope, landslide, tsunami, flood, storm-surge, and special alert-zone counts per property. Useful for disaster-risk portfolio screening and BCP analysis. Counts are a municipality-level approximation — the property itself may not sit inside the zone.

Parameters

Facilities owned by listed companies (from the "major facilities" section of annual securities reports), joined with co-located land price (commercial/residential/industrial), 2050 population forecast, and flood / landslide risk. Useful for regional-attribute analysis of a company's real-estate portfolio, unrealized gain/loss estimation, and disaster-risk assessment. Covers 267,164 facilities across 4,322 companies.

Parameters

Example Request

Transaction cases of non-REIT companies (real estate and fixed assets). Gemini structured-extracts the text blocks "plans for new / disposed facilities," "gain on sale of fixed assets," and "loss on sale of fixed assets" from EDINET annual securities reports. Searchable by company, property, location, transaction type, counterparty, and amount.

Parameters

Example Request

Note: Structured extraction by Gemini 2.5 Flash. Only transactions with an explicitly named property are retained (aggregate-only notes are skipped). Confidence is returned in the confidence field (high/medium/low).

Returns tenant concentration indicators aggregated from J-REIT disclosures: top-N GFA share (top1/3/5/10), Herfindahl-Hirschman Index (HHI, area-based and industry-based), number of industries, top industry share, area-weighted average remaining contract years, and contract-end peak year. Two scopes: reit (48 REITs) and property (1,535 properties). Useful for evaluating master-lease concentration, comparing tenant diversification across REITs, and forecasting contract renewal peaks.

Parameters

Request example

Note: Anonymous tenants (is_anonymous=true) are included in aggregates, but top1_named_tenant_name and named entries in top3_tenants are derived from non-anonymous rows only. Weight uses leased_area_m2 when available and falls back to leased_area_ratio_pct. Master-lease extraction is currently incomplete, so HHI is computed at the name-merged tenant level.

Returns one concentration record by property name. Exact match preferred, partial match as fallback.

Returns rent deviation (yen, %), z-score, and outlier flags for a J-REIT property, comparing its monthly rent per tsubo against the peer bucket use_type_norm × prefecture (sample ≥ 10). Falls back to national bucket when the prefecture sample is too small. Returns peer_basis='none' when neither has enough data.

use_type_norm categories

residential / office / retail / logistics / hotel / healthcare / other / _missing (use_type NULL in source)

Main fields

Request example

Note: Latest period only (time-series is not yet supported). Properties with peer_basis='none' (use_type NULL / other / parking) have all deviation indicators set to NULL. The logistics peer median can be distorted by partial ownership or shared-area inclusions — interpret with care. This is a simple peer-median comparison; the Phase 2 rent model v5 is not applied beyond residential. Asset-class-specific models are a planned extension.

Returns properties priced higher (outlier=high) or lower (outlier=low) than their peer distribution — useful for spotting rent-raise headroom or impairment candidates.

Borrowings-table contents parsed from each REIT's latest securities report and aggregated per REIT. Returns weighted-average interest rate, total outstanding balance, short/long-term split, weighted-average remaining years, share maturing within 1/3 years, and top-3 lender concentration. Useful for screening REITs exposed to rate-hike risk or high lender concentration. 65 REITs covered (latest period).

Parameters

Request example

Note: 65/69 REITs extracted (94%). 4 REITs have non-standard header layouts and are pending parser enhancement. Rate fill-rate 100%, remaining-years fill-rate 89%. Latest period only.

Per-loan details for a given REIT: lender bank, short/long-term category, begin/end balance, rate, repayment date, purpose, and collateral terms.