Skip to content
/ DoRA Public

DoRA: Domain-Based Self-Supervised Learning Framework for Low-Resource Real Estate Appraisal (CIKM-23)

Notifications You must be signed in to change notification settings

wwweiwei/DoRA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 

Repository files navigation

DoRA: Domain-Based Self-Supervised Learning Framework for Low-Resource Real Estate Appraisal (CIKM 2023)

Official code of the paper DoRA: Domain-Based Self-Supervised Learning Framework for Low-Resource Real Estate Appraisal. Paper link: http://arxiv.org/abs/2309.00855.

Overview

DoRA is pre-trained with an intra-sample geographic prediction as the pretext task based on the metadata of the real estate for equipping the real estate representations with prior domain knowledge. And, inter-sample contrastive learning is employed to generalize the representations to be robust for limited transactions of downstream tasks.

Model Framework

Model Training

python DoRA.py {building_type} {num_shot}

{building_type} = 'building', 'apartment', 'house'
{num_shot} = 1, 5, ...

Details of the Features

Real Estate Features

Feature Feature Type (#class) Example
City Name Category (21) Taipei city
Town Name Category (350) Ren’ai township
Parking spot Category (2) True/False (If the house includes a parking spot.)
Studio Binary True/False (If the area of the estate is smaller than 8 square meters.)
Details building type Category (5) Residential building (11 floors and above), Mansion (10 floors and below)
Main Purpose Category (1622) Electromechanical equipment space
Building materials Category (220) Rebar, Wood
Management organization Binary True/False
Type of parking space Category (3) Flat parking spot, Automated parking spot
Elevator Binary True/False
First-floor index Binary True/False (If the house is located on the first floor.)
Shop index Binary True/False (If the house is for shop use.)
Housing type Category (3) Building, Apartment, House
Village name Category (4650) Zhongshan village
Land use Category (19) Residential zone, Forestry land, Mining land
Land Use Designation Category (16) Type A building land, Class B building site
Land transfer area Numerical 30
Building transfer area Numerical 50
Number of bedrooms Numerical 2
Number of living rooms Numerical 1
Number of bathroom Numerical 3
Number of total rooms Numerical 5
Parking area Numerical 5
Main building area Numerical 100
Ancillary building area Numerical 10
Balcony area Numerical 5
House age Numerical 10 years
Number of land transaction Numerical 1
Number of building transaction Numerical 1
Number of parking space transactions Numerical 2
Building area without parking area Numerical 45
Single floor area Numerical 20
Floor area ratio (FAR) Numerical 10 (Derived by dividing the total area of the building by the total area of the parcel.)
Estate floor Numerical 5
Total floor Numerical 10
Latitude Numerical Horizontal lines that measure distance north or south of the equator
Longitude Numerical Vertical lines that measure east or west of the meridian in Greenwich, England.
Building coverage ratio Numerical 9
Park count flat Numerical 0

PoI Features

Generally, real estate with many YIMBY facilities often has a higher price since it implies the quality of living and the degree of transportation convenience. On the contrary, real estate with many NIMBY facilities may be likely to have a lower price since it indicates there may be some pollutant issues that cause a negative impact on living

Feature Feature Type (#class) Example
YIMBY_10 Numerical 2
YIMBY_50 Numerical 3
YIMBY_100 Numerical 3
YIMBY_250 Numerical 6
YIMBY_500 Numerical 7
YIMBY_1000 Numerical 13
YIMBY_5000 Numerical 28
YIMBY_10000 Numerical 52
NIMBY_10 Numerical 0
NIMBY_50 Numerical 0
NIMBY_100 Numerical 0
NIMBY_250 Numerical 1
NIMBY_500 Numerical 1
NIMBY_1000 Numerical 2
NIMBY_5000 Numerical 3
NIMBY_10000 Numerical 6

Economic and Geographical Features

Feature Feature Type (#class) Example
Land area per town Numerical 23.13 (km2)
Population density per town Numerical 23835 (#people/km2)
House price index per quarter Numerical 110
Unemployment rate per quarter Numerical 5%
Economic growth rate per quarter Numerical 3%
Lending rate per quarter Numerical 1.9%
Land transaction count per quarter Numerical 163796
Average land price index per quarter Numerical 101
Steel price index per quarter Numerical 1071

About

DoRA: Domain-Based Self-Supervised Learning Framework for Low-Resource Real Estate Appraisal (CIKM-23)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages