A mass-market appraisal of the English housing rental market using a diverse range of modelling techniques

Abstract

Introduction Mass appraisals in the rental housing market are far less common than those in the sales market. However, there is evidence for substantial growth in the rental market and this lack of insight hampers commercial organisations and local and national governments in understanding this market. Case description This case study uses data that are supplied from a property listings web site and are unique in their scale, with over 1.2 million rental property listings available over a 2 year period. The data is analysed in a large data institute using generalised linear regression, machine learning and a pseudo practitioner based approach. Discussion and evaluation The study should be seen as a practical guide for property professionals and academics wishing to undertake such appraisals and looking for guidance on the best methods to use. It also provides insight into the property characteristics which most influence rental listing price. Conclusions From the regression analysis, attributes that increase the rental listing price are: the number of rooms in the property, proximity to central London and to railway stations, being located in more affluent neighbourhoods and being close to local amenities and better performing schools. Of the machine learning algorithms used, the two tree based approaches were seen to outperform the regression based approaches. In terms of a simple measure of the median appraisal error, a practitioner based approach is seen to outperform the modelling approaches. A practical finding is that the application of sophisticated machine learning algorithms to big data is still a challenge for modern desktop PCs.

Parametric and Non-parametric Methods in Mass Appraisal on Poorly Developed Real Estate Markets*

“…An example of such work is the paper of (Park and Bae, 2015), in which the effectiveness of real estate price predictions in Fairfax County, Virginia, was analysed. English housing rental market was subjected to mass appraisal using generalized linear regression, machine learning and a pseudo practitioner based approach (Clark and Lomax, 2018). Apart from the conclusions regarding the fact that machine learning models proved to be superior to multiple regression, the authors argue that the use of machine learning is computationally demanding, which was also confirmed in this study.…”

A Fully Automated Adjustment of Ensemble Methods in Machine Learning for Modeling Complex Real Estate Systems

“…e results show a better behaviour of ensemble of M5 model trees with a better behaviour of bagging unpruned decision trees, with a mean relative error of 15.25%. Similar results with a median percentage error of 15.11% and 13.18% are obtained for the English private rental market using gradient boost [34] and Cubist [35], respectively, by Clark and Lomax [36]. Graczyk et al [37] use six machine learning algorithms: multilayer perceptron (MLP); radial basis function neural network for regression problems (RBF); pruned model tree (M5P); M5Rules (M5R); linear regression model (LRM); and NU-support vector machine (SVM) for the three ensemble methods of additive regression (an implementation of boosting in WEKA), bagging, and stacking, in Waikato Environment for Knowledge Analysis (WEKA).…”

“…ese same aspects are highlighted in [55,56] in which the authors point out that web prices offer a valuable opportunity for statistical analysis due to the constant generation of information, their accessibility, and availability as well as there being little notable differences compared to offline prices. Within real estate, applications developed using web data are used byÖzsoy and Şahin [23] in Istanbul; Del Cacho [14] in Madrid; Larraz and Larraz and Población [57,58] in Spain; Pow et al [17] in Montreal; Larraz and Población [59] in Czech Republic; Nguyen [25] in the United States; Clark and Lomax [36] in England; Pérez-Rave et al [20] in Colombia; or Neloy et al [44] in Bangladesh.…”

Impact of rail transit station proximity to commercial property prices: utilizing big data in urban real estate

“…Asking price using real estate marketplaces is categorized as a part of volunteered geographic information (VGI) uploaded by the users in voluntary-basis tagged with geolocation [18]. The asking price using this approach is argued as one of the sources of big data characterized by its large samples with a variety of contents both in a structured and unstructured manner, and the speed of data processing [33]. These types of datasets may promote better planning in urban and transportation areas by leaving subjectivity in data collection from personal interests.…”