jcrainic2/causal-real-estate

Causal Real Estate数据集是一个多城市房产级别的数据集，包含波士顿、纽约（曼哈顿和布鲁克林）和旧金山的556,636处房产信息。数据集结合了房产评估、人口普查人口统计、地理编码的犯罪事件和便利设施点。此外，还提供了约3,000处房产的句子转换器嵌入数据。数据集主要用于研究房地产估值中的空间混杂问题，并支持因果推断、城市分析等任务。数据集的结构包括房产属性、位置信息、人口普查数据、犯罪记录、便利设施和微地理信息等。The Causal Real Estate dataset is a multi-city parcel-level dataset containing information on 556,636 parcels across Boston, New York (Manhattan + Brooklyn), and San Francisco. The dataset combines property assessments, census demographics, geocoded crime incidents, and points-of-interest amenities. Additionally, it includes sentence-transformer embeddings for a subset of ~3,000 parcels. The dataset is primarily used for studying spatial confounding in real estate valuation and supports tasks such as causal inference and urban analytics. The dataset structure includes property attributes, location information, census data, crime records, amenities, and micro-geography details.