RRE-GTC: A Geo-Temporal Cluster Dataset for Real-Estate PriceEstimation and Market-Aware Search

Real-estate platforms represent a complex Information Retrieval (IR) environment where user relevance depends on high-dimensional intersections of location, time, and physical attributes. However, research into Geo-spatial Information Retrieval (GIR) and location-based recommendations is currently impeded by a lack of open, high-quality datasets that capture both the temporal evolution of markets and granular accessibility signals. To bridge this gap, we introduce RRE-GTC (Ru-Real-Estate Geo-Temporal Clusters), an open resource designed to benchmark ranking, pricing, and recommendation algorithms in dynamic spatial contexts. Derived from a massive collection of apartment listings in major Russian cities, RRE-GTC aggregates 456,182 geo-temporal clusters, offering a privacy-preserving alternative to releasing raw listing-level data. The dataset provides rich and distinct features: (i) geo-temporal price dynamics (stratified percentiles over time), (ii) detailed item metadata (building stock, supply volume), and (iii) transport accessibility signals (e.g., proximity, transport density). Uniquely, the resource includes a ”transport-isolated” price signal derived from a linear attribution model, enabling novel research into attribute-based ranking and explainable search (e.g., separating ”location value” from ”intrinsic quality”). Released under the Apache-2.0 license and hosted on Hugging Face, RRE-GTC lowers the barrier for experimenting with content-based recommendation and retrieval, spatiotemporal ranking, and market-aware search filtration.