UrbanScore: A Real-Time Personalized Liveability Analytics Platform

This paper introduces UrbanScore — a real-time web platform that computes a personalized liveability score for any urban address. The system fuses five data streams: (i) address geocoding via Nominatim, (ii) facility extraction from OpenStreetMap through Overpass QL, (iii) segment-level traffic metrics from TomTom Flow v10, (iv) hourly air-quality readings from OpenWeatherMap, and (v) user-declared preference profiles, all persisted in an Oracle 19c relational store. Six sub-scores (air, traffic, lifestyle, education, metro access, surface transport) are derived, adaptively weighted and combined; an OpenAI large-language model then converts the numeric results into concise, user-friendly explanations. A pilot deployment covering the 226 km2 metropolitan area of Bucharest evaluated 3 450 unique addresses over four weeks. Median end-to-end latency was 2.1 s (p95 = 2.9 s), meeting the !3 s non-functional requirement. Aggregate scores ranged from 34 to 92 (mean 68, SD 11), with high-scoring clusters along metro corridors that pair abundant green space with PM2.5 levels below 35 μg m−3. A detailed case study of the Tineretului district produced an overall score of 91/100 and demonstrated how the narrative layer guides users toward comparable neighborhoods. Limitations include dependence on third-party API uptime, spatial bias toward well mapped OSM regions and the absence of noise and crime layers, cited by 18 % of survey participants as a desired enhancement. Overall, the results show that open geodata, commercial mobility feeds and conversational AI can be integrated into a performant, explainable decision-support tool that places “liveability analytics†in the hands of every house-hunter, commuter and city planner.