Recommender Systems in the Real Estate Market—A Survey

Abstract

The shift to e-commerce has changed many business areas. Real estate is one of the applications that has been affected by this modern technological wave. Recommender systems are intelligent models that assist users of real estate platforms in finding the best possible properties that fulfill their needs. However, the recommendation task is substantially more challenging in the real estate domain due to the many domain-specific limitations that impair typical recommender systems. For instance, real estate recommender systems usually face the clod-start problem where there are no historical logs for new users or new items, and the recommender system should provide recommendations for these new entities. Therefore, the recommender systems in the real estate market are different and substantially less studied than in other domains. In this article, we aim at providing a comprehensive and systematic literature review on applications of recommender systems in the real estate market. We evaluate a set of research articles (13 journal and 13 conference papers) which represent the majority of research and commercial solutions proposed in the field of real estate recommender systems. These papers have been reviewed and categorized based on their methodological approaches, the main challenges that they addressed, and their evaluation procedures. Based on these categorizations, we outlined some possible directions for future research.

Applications, Opportunities and Challenges of AI in the Real Estate Sector

“…As an application of AI in the RE sector, a study was conducted regarding the implementation of a recommendation system in a website that shares advertisements from different RE agencies [88], [89]. Three different initial models were testedcontent-based recommendation, incremental collaborative filtering and frequency and recency on view item recommendation -and throw an A/B testing the authors conclude that with their system they can promote bigger conversion rates, even demonstrating a 250% increase in the performance of the industry professionals [90].…”

Changes in warehouse spatial patterns and rental prices: Are they related? Exploring the case of US metropolitan areas

“…Therefore, we have performed an automated web data extraction method: web scraping. Individual scripts were developed to scrape the warehouse information from the Html file and assemble one dataset for each metropolitan area (Ge et al, 2021;Gharahighehi et al, 2021;Jiao et al, 2021;Luo and He, 2021;Pineda-Jaramillo and Pineda-Jaramillo, 2021;Ploessl et al, 2021). This method was mainly composed of seven steps for this paper: (i) find the URL where the data is published; (ii) inspect the webpage to find the data from its source code; (iii) build the prototype code, which was written in R language and using the rvest package and intended to extract and prepare the data; (iv) generalize the code considering functions, loops and debugging and run the code alternating among US metropolitan areas; (v) store the data as an organized data frame; (vi) check and clean the gathered data; (vii) geocode the information on each warehouse.…”

An Adaptive Hybrid Active Learning Strategy with Free Ratings in Collaborative Filtering

“…Sparsity is one of the major issues in CF. There are two main approaches in the literature to address sparsity: generating artificial user logs and rating elicitation [5]. Chonwiharnphan et al [3] proposed a generative model to generate artificial user logs.…”

“…The sparsity issue arises as users interact with a small fraction of the item catalog and therefore all the other possible interactions are missing. To address this issue, there are two main approaches: generating artificial logs via generative models, as suggested in [5], or requesting real users to rate specific items, as performed by active learning (AL) strategies [4]. Despite being vastly studied in other tasks, generating artificial data in RSs is still in its initial steps [22].…”

“…Each user can have ratings of different items in the test set 4. We used 40 for k 5. We used 291 latent factors,1501 iterations, 0.01834 for learning rate and 0.01467 for regularization.…”