HK-BFETD: Hong Kong Building Function and Energy Typology Dataset

HK-BFETD is a building-level bilingual (English_Chinese) dataset for Hong Kong, developed to support urban building energy modeling (UBEM) and related spatial energy analyses. It provides EMSD-aligned functional typologies, mixed-use proportional vectors, and geospatial building units that enable consistent linkage between official sectoral energy statistics and building-level spatial records. This release (v1.1.0) includes:1) HK_UBEM_Buildings_Public_v1_1.csv2) HK_UBEM_Buildings_Public_v1_1.geojson The tabular product contains 341,153 building records and 30 fields, including footprint/floor attributes, UBEM main and sub classes, mixed-use proportions, classification provenance, and probabilistic outputs for uncertainty-aware allocation. Main-class counts in this release are:- Residential_住宅类别: 251,901- Commercial_商业类别: 49,129- Mixed-use_混合用途: 31,395- Non-assessed_非评估类别: 5,373- Industrial_工业类别: 3,355 Usage scope:HK-BFETD is designed as a spatial typology and allocation dataset for benchmark-informed urban energy applications. It is suitable for city-scale stock characterization, mixed-use allocation, sectoral aggregation, scenario construction, and policy-oriented comparative analysis. For mixed-use buildings, UBEM_Mixed_Proportions can be used as initial allocation weights when distributing area- or emission-related quantities across functional sectors. Interpretation note:Probability vectors represent epistemic uncertainty and should be interpreted as probabilistic allocation weights rather than direct measurements. The dataset does not provide building-specific measured energy consumption, conditioned floor area, HVAC system characteristics, or detailed operational schedules; therefore, detailed building-by-building simulation requires additional project-specific assumptions and external data. Traceability:Classification_Source and related provenance fields support source-aware filtering and uncertainty management. Deterministic preprocessing and rule-based stages are script-traceable, while AI-assisted stages are documented through prompts, thresholds, logs, and workflow metadata. Exact byte-identical replay of API-mediated outputs may depend on external model availability and service state. Data provenance note:Some raw source datasets used in preprocessing (e.g., official or third-party source layers) are governed by their own licenses and are therefore not redistributed in this Zenodo record. This record publishes the release-ready derived products for reuse and citation. The code repository associated with the workflow is available at https://github.com/yizheng-hub/HKBFETD.git.