EzModel

PROPOSAL

POSTER

APPLICATION

RESEARCH PAPER

Project Description

In recent decades, there has been a surging interest in modeling housing prices among economists, planners, and policymakers due to the significant role of properties in household wealth and national economy. In Singapore, public housing accommodates more than 80% of its citizen and citizens either choose to buy a new Housing Development Board (HDB) flat or purchase a HDB resale flat, second-hand flats with less than 99 years left on the lease.

Our project will focus on modelling the HDB resale flat prices which are shaped by market forces. As many previous hedonic pricing models that uses linear regression fails to take into account spatial variations among the observations in the local surroundings, our project will be building a modeling tool based on the geographically weighted regression (GWR) model to analyse the effects of spatial variation on housing prices. Our application will provide users with the option of using a mixed geographically weighted model as well as allows users to add new spatial attributes into the GWR model that will be computed on the go based on the dataset uploaded. We hope that this modelling tool will help users more accurately investigate the impact of variables on HDB resale flat prices in Singapore.

Project Motivation

In recent decades, there has been a surging interest in modeling housing prices among economists, planners, and policymakers due to the significant role of properties in household wealth and national economy. In Singapore, public housing accommodates more than 80% of its citizen and citizens either choose to buy a new Housing Development Board (HDB) flat or purchase a HDB resale flat, second-hand flats with less than 99 years left on the lease. Our project will focus on modelling the HDB resale flat prices which are shaped by market forces.

In many existing hedonic housing prices models, linear regression is used to identify the significance of different variables (number of rooms, number of years left on the lease and distance from the nearest amenities etc.) on the house prices. However, these models fail to take into account the spatial variation in the nearby surroundings of these different resale units such as the proximity to shopping malls, number of MRT stations and healthcare facilities in the vicinity.

Geographically weighted regression (GWR) is a spatial analysis technique that overcomes this limitation by taking into account spatial autocorrelations among the observations in surrounding locations by allowing for spatial nonstationarity in the linear regression coefficients for each observation location. In this project, we will build a modeling tool that allows users to explore the impact of these spatial variations on HDB resale prices through a GWR model. To factor in the combination of both local and global variables, we also includes the option of using a mixed (semiparametric) GWR model.

Also, to provide greater flexibility for users to choose the certain desired spatial attributes that they would like to analyse, instead of fixing the datasets to be used for the model, we will allow users to upload datasets (i.e. school locations, hospital locations) and the tool will immediately compute new spatial variables to allow users to include in the model for analysis.This tool thus seeks to help users to accurately model the impact of spatial variables on the price of the HDB resale units.

Data Sources

Literature Review

Approach

Project Timeline

Challenges

Tools & Technology

Future Extensions