1# RECORD 05c8066a4f1ebf0565508520c4002c90671161c42# TITLE: Algorithm-Driven Hedonic Real Estate Pricing – An Explainable AI Approach3# YEAR: 2024.0 VENUE: Real Estate Management and Valuation DOI: 10.2478/remav-2025-00034# AUTHORS: S. Tarasov; Bartłomiej Dessoulavy-Śliwiński5# PAGES: 13 CANDIDATE TABLE PAGES (1-based): []6# ABSTRACT: Abstract Data-driven machine learning algorithms triggered a fundamental change in hedonic real estate pricing. However, their adaptive nonparametric structure makes inference and out-ofsample prediction challenging. This study introduces an explainable approach to interpreting machine learning predictions, which has not been done before in the local market context. Specifically, Random Forest and Extreme Gradient Boosting models are developed for residential real estate price prediction in Warsaw in 2021 on 10,827 property transactions. Model-agnostic Explainable Artificial Intelligence (XAI) methods are then used to investigate the black box decision making. The results show the practicability of applying XAI frameworks in the real estate market context to decode the rationale behind data-driven algorithms. Information about the relationships between input variables is extracted in greater detail. Accurate, reliable and transparent real estate valuation support tools can offer substantial advantages to participants in the real estate market, including banks, insurers, pension and sovereign wealth funds, as well public authorities and private individuals.789===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) =====1011 12 13REAL ESTATE MANAGEMENT AND VALUATION - vol. 33, no. 1, 2025 14eISSN: 2300-5289 | © 2024 The Author(s) | Article under the CC BY 4.0 license 1522 16ALGORITHM-DRIVEN HEDONIC REAL ESTATE PRICING – AN 17EXPLAINABLE AI APPROACH 18 19Stanislav Tarasov1, Bartłomiej Dessoulavy-Śliwiński2* 20 211 Faculty of Economic Sciences, University of Warsaw, Długa 44/50, 00-241 Warsaw, Poland, e-mail: 22swtarasow@gmail.com, ORCID: 0000-0001-7018-9847 232 Department of Management and Information Technology, Faculty of Economic Sciences, University of Warsaw, Długa 2444/50, 00-241 Warsaw, Poland, e-mail: bsliwinski@wne.uw.edu.pl, ORCID: 0000-0002-4395-9815 25* Corresponding author 26 27ARTICLE INFO 28ABSTRACT 29Keywords: 30Data-driven machine learning algorithms triggered a fundamental change in hedonic real 31estate pricing. However, their adaptive nonparametric structure makes inference and out-of-32sample prediction challenging. This study introduces an explainable approach to interpreting 33machine learning predictions, which has not been done before in the local market context. 34Specifically, Random Forest and Extreme Gradient Boosting models are developed for 35residential real estate price prediction in Warsaw in 2021 on 10,827 property transactions. 36Model-agnostic Explainable Artificial Intelligence (XAI) methods are then used to investigate 37the black box decision making. The results show the practicability of applying XAI frameworks 38in the real estate market context to decode the rationale behind data-driven algorithms. 39Information about the relationships between input variables is extracted in greater detail. 40Accurate, reliable and transparent real estate valuation support tools can offer substantial 41advantages to participants in the real estate market, including banks, insurers, pension and 42sovereign wealth funds, as well public authorities and private individuals. 43residential real estate, machine 44learning, 45explainable 46artificial 47intelligence, 48mass 49appraisal, 50automated valuation models 51JEL Classification: 52C14, C18, R39 53 54Citation: 55 56Tarasov, S. & Dessoulavy-Śliwiński, B. (2025). Algorithm-driven hedonic real estate pricing - 57An explainable AI approach. Real Estate Management and Valuation, 33(1), 22-34. 58https://doi.org/10.2478/remav-2025-0003 59 601. Introduction 61Real estate is the world’s biggest asset class, and 62residential real estate is a major asset for many 63individuals. Hence, property price prediction holds 64significant importance and increasingly attracts 65attention of researchers due to a variety of factors, 66including 67investment 68decisions, 69buy 70and 71sell 72transactions, tax considerations, or real estate 73investment trusts’ activity (Arribas et al., 2016). Due to 74market and regulatory changes in recent years, real 75estate is appraised more frequently than ever before, 76and inaccurate valuation can lead to various 77complications affecting property owners, buyers, and 78financing providers. Hence, there is a strong demand 79for accurate, speedy, and dependable support tools 80for real estate valuation. As a result, automated 81valuation model (AVM) development has been 82steadily on the rise in the industry and in academic 83research. 84 85On the one hand, the subject of predicting real 86estate prices is well recognized in the literature. 87Studies on this topic in the field of applied 88econometrics often employ conventional linear 89regression 90methods 91along 92with 93their 94spatial 95extensions, as well as ML-based models. Research has 96indicated varying results when comparing the 97performance of ML and spatial models, with some 98studies demonstrating superior results for the former 99(e.g., Rico-Juan & Taltavull de La Paz, 2021), and 100others presenting contrasting findings, in favor of the 101latter (e.g., Przekop, 2022). Li (2022), in their 102experimental study of extracting spatial effects from 103an ML model using a local interpretation method, 104shows that “locally interpreted machine learning 105models are good alternatives to spatial statistical 106models and perform better when complex spatial and 107non-spatial effects (e.g. non-linearities, interactions) 108coexist and are unknown”. On the other hand, 109following the reasoning of Angrist and Pischke 110(2010), “empirical evidence on any given causal effect 111