Prototype-based learning for real estate valuation - a machine learning model that explains prices

Aug 3, 2026

Upload by Aurelia

Systematic prediction of real estate prices underpins firm revenue management and supports societal and policy decisions. Prior work relies on ordinary least squares or other machine learning models that may achieve strong accuracy but often diverge from how humans reason about prices. Humans commonly use direct comparison by matching a target property to similar ones, a principle rarely encoded in standard ML pipelines. This article introduces prototype-based learning tailored to valuation, capturing non-linearity, optimizing a chosen loss, providing explainability, and explicitly encoding direct comparison; experiments show equal or better predictive accuracy and efficient summarization via representative prototypes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages25

Reading Time1 h 3 min

AccessFree to download

Abstract

Introduction

Real estate valuation and stakeholders

Hedonic pricing and OLS

Criticisms of OLS

Machine learning approaches (RFs, neural networks)

Need to encode direct comparison

Annals of Operations Research (2025) 344:287–311 [https://doi.org/10.1007/s10479-024-06273-1](https://doi.org/10.1007/s10479-024-06273-1) ORIGINAL RESEARCH Prototype-based learning for real estate valuation: a machine learning model that explains prices Jose A. Rodriguez-Serrano1 Received: 22 December 2023 / Accepted: 3 September 2024 / Published online: 23 September 2024 © The Author(s) 2024 Abstract The systematic prediction of real estate prices is a foundational block in the operations of many ﬁrms and has individual, societal and policy implications. In the past, a vast amount of works have used common statistical models such as ordinary least squares or machine learning approaches. While these approaches yield good predictive accuracy, most models work very differently from the human intuition in understanding real estate prices. Usually, humans apply a criterion known as “direct comparison”, whereby the property to be valued is explicitly compared with similar properties. This trait is frequently ignored when applying machine learning to real estate valuation. In this article, we propose a model based on a methodology called prototype-based learning, that to our knowledge has never been applied to real estate valuation. The model has four crucial characteristics: (a) it is able to capture non-linear relations between price and the input variables,(b) it is a parametric model able to optimize any loss function of interest,(c) it has some degree of explainability, and, more importantly, (d) it encodes the notion of direct comparison. None of the past approaches for real estate prediction comply with these four characteristics simultaneously. The experimental validation indicates that, in terms of predictive accuracy, the proposed model is better or on par to other machine learning based approaches. An interesting advantage of this method is the ability to summarize a dataset of real estate prices into a few “prototypes”, a set of the most representative properties. Keywords Real estate valuation · Non-linear regression · Machine learning · Prototype-based models 1 Introduction Real estate valuation (Pagourtzi et al., 2003; d’Amato and Kauko, 2017) refers to estimating the price of a property with accuracy on the basis of the current market, and impacts a wide range of stakeholders across industry and society, such as individuals and organizations involved in real estate transactions, or policymakers seeking to manage urban development and housing policies effectively. B Jose A. Rodriguez-Serrano [joseantonio.rodriguez15@esade.edu](joseantonio.rodriguez15@esade.edu) 1 Department of Operations, Innovation and Data Sciences, Universitat Ramon Llull Esade, Barcelona, Spain From an operations research standpoint, the ability of ﬁrms to predict real estate prices with precision is crucial for revenue management (Tchuente and Nyawa, 2022), portfolio optimization (Amédée-Manesme and Barthélémy, 2018; Kok et al., 2017), or risk management (Doumpos et al., 2021) and serves as a foundational step which impacts the operations of speciﬁc sectors, such as insurance or market research. The progress in digitization and increased data accessibility has enabled the systematic and large-scale estimation of property prices, known as mass valuation (McCluskey et al., 2013; Kok et al., 2017), often involving the use of statistical or machine learning techniques. According to Alexandridis et al. (2019),“automatic mass valuation can reduce operational costs since it is inexpensive and can be performed in a regular basis”. A common approach to mass valuation of real estate properties is hedonic pricing(Malpezzi,2003). Hedonic models decompose the utility of agoodin the sum ofutilities of its characteristics independently. An iconic example of a hedonic price model is OLS regression (James et al., 2013), wherein the unit value assigned to each characteristic corresponds to the regression coefﬁcient. OLS has been widely

FAQs

What is the main limitation of many existing machine learning models for real estate valuation?

They often achieve good predictive accuracy but do not reflect the human “direct comparison” criterion used to reason about prices.

How does the proposed prototype-based learning model differ from prior approaches?

It encodes direct comparison while also capturing non-linear relations, optimizing an arbitrary loss in a parametric framework, and offering a degree of explainability.

What benefits does prototype-based learning provide beyond prediction accuracy?

It summarizes a real estate dataset into a small set of representative properties called prototypes.

Aug 3, 2026

Upload by Aurelia