Scholarlink

This research examines the real estate market in Argentina, aiming to contribute socially by providing a key tool for the accurate valuation of properties. This, in turn, promotes efficiency in resource allocation, enhances transparency, and fosters competition within the Argentine real estate sector. Furthermore, improving price prediction through machine learning constitutes a fundamental tool for the adoption of more effective housing policies. With over 63,000 records of apartment sales collected online and merged with geographical data, we applied hedonic regression and machine learning algorithms to predict sale prices. The most accurate method was XGBoost. In addition to focusing on price prediction, the study also emphasizes model interpretability. For this reason, we applied model-agnostic interpretability methods, including Partial Dependence Plots (PDP), Individual Conditional Expectation (ICE), and SHapley Additive exPlanations (SHAP). These methods provide transparency regarding which features are relevant for valuation, thereby complementing traditional approaches such as hedonic regression.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Real Estate Market in Buenos Aires (Argentina): A Hedonic and Interpretable Machine Learning Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Emiliano Gutiérrez,
  • Laura Lanzarini,
  • Fernando Delbianco,
  • Rocío Cecchini

摘要

This research examines the real estate market in Argentina, aiming to contribute socially by providing a key tool for the accurate valuation of properties. This, in turn, promotes efficiency in resource allocation, enhances transparency, and fosters competition within the Argentine real estate sector. Furthermore, improving price prediction through machine learning constitutes a fundamental tool for the adoption of more effective housing policies. With over 63,000 records of apartment sales collected online and merged with geographical data, we applied hedonic regression and machine learning algorithms to predict sale prices. The most accurate method was XGBoost. In addition to focusing on price prediction, the study also emphasizes model interpretability. For this reason, we applied model-agnostic interpretability methods, including Partial Dependence Plots (PDP), Individual Conditional Expectation (ICE), and SHapley Additive exPlanations (SHAP). These methods provide transparency regarding which features are relevant for valuation, thereby complementing traditional approaches such as hedonic regression.