The Importance of Economic Variables on London Real Estate Market: A Random Forest Approach

Gabriella Piscopo,Susanna Levantesi

doi:10.3390/risks8040112

Abstract

This paper follows the recent literature on real estate price prediction and proposes to take advantage of machine learning techniques to better explain which variables are more important in describing the real estate market evolution. We apply the random forest algorithm on London real estate data and analyze the local variables that influence the interaction between housing demand, supply and price. The variables choice is based on an urban point of view, where the main force driving the market is the interaction between local factors like population growth, net migration, new buildings and net supply.

Highlights

Urban economy and real estate markets are two interconnected fields of research
The point of view offered is different and complementary with respect to the literature on the field, which considers features attaining the buildings like size and location, and is based on an urban perspective to explain the evolution of the local real estate market
The use of random forest (RF) in small datasets is common among data scientists as the bootstrapping, on which RF is based, allows the algorithm to perform well anyway

Summary

Introduction

Urban economy and real estate markets are two interconnected fields of research. They overlap in so far as the real estate price evolution is analyzed under an urban approach: it has been shown that less than 8% of the variation in price levels across cities can be accounted for by national effects. This paper follows the recent developments of the literature on real estate and proposes to take advantage of the random forest algorithm to better explain which variables have more importance in describing the evolution of the house price following an urban approach. To this aim, we focus on a given city, and analyze the local variables that influences the interaction between housing demand and supply and the price.

The Model

Regression Tree Architecture

Random Forest

Variable Importance

Case Study

Data Description

Regression Model and Main Statistics

RF Estimation of AHP

Sensitivity to Predictor

RF Predictive Performance and Comparison with GLM

Findings

Conclusions