A 2023 Shenzhen land valuation modeling dataset containing data on accessibility, rent, housing prices, and land transaction samples.

This dataset supports land valuation modeling for Shenzhen, China, in the 2023 study context. It contains three geospatial data layers and one accompanying data dictionary. The main layer is a 300 m x 300 m polygon grid (`predictor_result_grid.shp`) with 21,362 grid cells and 19 non-geometry attributes describing predictor variables and modeled land-price values. Two point layers are also included: `model_training_samples.shp` with 13 records and 30 non-geometry attributes, and `model_validation_samples.shp` with 8 records and 30 non-geometry attributes. The polygon layer stores environmental, socioeconomic, accessibility, and land-price variables, including rental price, housing price, NDVI, a population-related predictor, distance to the central business district, and walking time in minutes to the nearest primary school, junior secondary school, hospital, shopping mall, supermarket, commercial building, subway station, and park. The sample layers store land transaction and planning-control attributes used in model training and validation. A machine-readable and human-readable field dictionary is provided in `data_inventory.csv` and `data_inventory.md`.The walking-time variables were derived using the Amap (Gaode Map) API.The land transaction dataset was obtained from the publicly available records of the Shenzhen Public Resources Trading Center (https://new.szggzy.com/jyfw/list.html?id=jyfwtdky).The released files were standardized from project Shapefiles supplied in the working directory. Public-release processing was carried out in Python using GeoPandas and Pyogrio to preserve geometries and attribute values while converting non-English field names into compact English names suitable for public dissemination. No geometry simplification, spatial resampling, or missing-value imputation was performed during this release-preparation step. Because the ESRI Shapefile format restricts field names to 10 characters, the published files use abbreviated ASCII field names, while the full field meanings, units, and interpretation notes are documented in the accompanying inventory files. The main grid layer is stored in EPSG:3857 and represents a spatial resolution of 300 m x 300 m (0.09 km2 per grid cell). The training and validation point layers are stored in EPSG:4326. The grid layer also includes longitude and latitude attributes to support geographic reference.Regarding data completeness, the polygon grid layer contains no missing attribute values. No explicit per-feature uncertainty field or numerical error range is included in the release. Potential uncertainty may arise from the quality of the original source data, accessibility estimation procedures, land-price standardization steps, and modeling assumptions in the associated study. The spatial files are distributed in standard ESRI Shapefile format with the required `.shp`, `.shx`, `.dbf`, `.prj`, and `.cpg` sidecar files and can be opened in common GIS software such as QGIS, ArcGIS Pro, or Python GeoPandas. The approximate package sizes are 12.13 MB for the polygon grid layer, 0.02 MB for the training sample layer, and 0.02 MB for the validation sample layer.