Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework

Yihan Guo, Shan Lin, Xiao Ma, Jay Bal, Chang tsun Li

Research output: Book chapter/Published conference paperConference paper

Abstract

Most existing real estate appraisal methods focus on building accuracy and reliable models from a given dataset but pay little attention to the extensibility of their trained model. As different cities usually contain a different set of location features (district names, apartment names), most existing mass appraisal methods have to train a new model from scratch for different cities or regions. As a result, these approaches require massive data collection for each city and the total training time for a multi-city property appraisal system will be extremely long. Besides, some small cities may not have enough data for training a robust appraisal model. To overcome these limitations, we develop a novel Homogeneous Feature Transfer and Heterogeneous Location Fine-tuning (HFT+HLF) cross-city property appraisal framework. By transferring partial neural network learning from a source city and fine-tuning on the small amount of location information of a target city, our semi-supervised model can achieve similar or even superior performance compared to a fully supervised Artificial neural network (ANN) method.

Original languageEnglish
Title of host publicationData Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers
EditorsYanchang Zhao, Graco Warwick, David Stirling, Chang-Tsun Li, Yun Sing Koh, Rafiqul Islam, Zahidul Islam
PublisherSpringer-Verlag London Ltd.
Pages161-174
Number of pages14
ISBN (Print)9789811366604
DOIs
Publication statusPublished - 01 Jan 2019
Event16th Australasian Conference on Data Mining, AusDM 2018 - Bathurst, Australia
Duration: 28 Nov 201830 Nov 2018

Publication series

NameCommunications in Computer and Information Science
Volume996
ISSN (Print)1865-0929

Conference

Conference16th Australasian Conference on Data Mining, AusDM 2018
CountryAustralia
CityBathurst
Period28/11/1830/11/18

Fingerprint

Tuning
Neural networks
Model
Artificial Neural Network
Framework
Neural Networks
Partial
Target
Training

Cite this

Guo, Y., Lin, S., Ma, X., Bal, J., & Li, C. T. (2019). Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework. In Y. Zhao, G. Warwick, D. Stirling, C-T. Li, Y. S. Koh, R. Islam, & Z. Islam (Eds.), Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers (pp. 161-174). (Communications in Computer and Information Science; Vol. 996). Springer-Verlag London Ltd.. https://doi.org/10.1007/978-981-13-6661-1_13
Guo, Yihan ; Lin, Shan ; Ma, Xiao ; Bal, Jay ; Li, Chang tsun. / Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework. Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers. editor / Yanchang Zhao ; Graco Warwick ; David Stirling ; Chang-Tsun Li ; Yun Sing Koh ; Rafiqul Islam ; Zahidul Islam. Springer-Verlag London Ltd., 2019. pp. 161-174 (Communications in Computer and Information Science).
@inproceedings{07ca12ed02fc43618ecf259280defcfe,
title = "Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework",
abstract = "Most existing real estate appraisal methods focus on building accuracy and reliable models from a given dataset but pay little attention to the extensibility of their trained model. As different cities usually contain a different set of location features (district names, apartment names), most existing mass appraisal methods have to train a new model from scratch for different cities or regions. As a result, these approaches require massive data collection for each city and the total training time for a multi-city property appraisal system will be extremely long. Besides, some small cities may not have enough data for training a robust appraisal model. To overcome these limitations, we develop a novel Homogeneous Feature Transfer and Heterogeneous Location Fine-tuning (HFT+HLF) cross-city property appraisal framework. By transferring partial neural network learning from a source city and fine-tuning on the small amount of location information of a target city, our semi-supervised model can achieve similar or even superior performance compared to a fully supervised Artificial neural network (ANN) method.",
keywords = "Mass appraisal, Property valuation, Transfer learning",
author = "Yihan Guo and Shan Lin and Xiao Ma and Jay Bal and Li, {Chang tsun}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-981-13-6661-1_13",
language = "English",
isbn = "9789811366604",
series = "Communications in Computer and Information Science",
publisher = "Springer-Verlag London Ltd.",
pages = "161--174",
editor = "Yanchang Zhao and Graco Warwick and David Stirling and Chang-Tsun Li and Koh, {Yun Sing} and Rafiqul Islam and Zahidul Islam",
booktitle = "Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers",
address = "Germany",

}

Guo, Y, Lin, S, Ma, X, Bal, J & Li, CT 2019, Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework. in Y Zhao, G Warwick, D Stirling, C-T Li, YS Koh, R Islam & Z Islam (eds), Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers. Communications in Computer and Information Science, vol. 996, Springer-Verlag London Ltd., pp. 161-174, 16th Australasian Conference on Data Mining, AusDM 2018, Bathurst, Australia, 28/11/18. https://doi.org/10.1007/978-981-13-6661-1_13

Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework. / Guo, Yihan; Lin, Shan; Ma, Xiao; Bal, Jay; Li, Chang tsun.

Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers. ed. / Yanchang Zhao; Graco Warwick; David Stirling; Chang-Tsun Li; Yun Sing Koh; Rafiqul Islam; Zahidul Islam. Springer-Verlag London Ltd., 2019. p. 161-174 (Communications in Computer and Information Science; Vol. 996).

Research output: Book chapter/Published conference paperConference paper

TY - GEN

T1 - Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework

AU - Guo, Yihan

AU - Lin, Shan

AU - Ma, Xiao

AU - Bal, Jay

AU - Li, Chang tsun

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Most existing real estate appraisal methods focus on building accuracy and reliable models from a given dataset but pay little attention to the extensibility of their trained model. As different cities usually contain a different set of location features (district names, apartment names), most existing mass appraisal methods have to train a new model from scratch for different cities or regions. As a result, these approaches require massive data collection for each city and the total training time for a multi-city property appraisal system will be extremely long. Besides, some small cities may not have enough data for training a robust appraisal model. To overcome these limitations, we develop a novel Homogeneous Feature Transfer and Heterogeneous Location Fine-tuning (HFT+HLF) cross-city property appraisal framework. By transferring partial neural network learning from a source city and fine-tuning on the small amount of location information of a target city, our semi-supervised model can achieve similar or even superior performance compared to a fully supervised Artificial neural network (ANN) method.

AB - Most existing real estate appraisal methods focus on building accuracy and reliable models from a given dataset but pay little attention to the extensibility of their trained model. As different cities usually contain a different set of location features (district names, apartment names), most existing mass appraisal methods have to train a new model from scratch for different cities or regions. As a result, these approaches require massive data collection for each city and the total training time for a multi-city property appraisal system will be extremely long. Besides, some small cities may not have enough data for training a robust appraisal model. To overcome these limitations, we develop a novel Homogeneous Feature Transfer and Heterogeneous Location Fine-tuning (HFT+HLF) cross-city property appraisal framework. By transferring partial neural network learning from a source city and fine-tuning on the small amount of location information of a target city, our semi-supervised model can achieve similar or even superior performance compared to a fully supervised Artificial neural network (ANN) method.

KW - Mass appraisal

KW - Property valuation

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85063474823&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063474823&partnerID=8YFLogxK

U2 - 10.1007/978-981-13-6661-1_13

DO - 10.1007/978-981-13-6661-1_13

M3 - Conference paper

SN - 9789811366604

T3 - Communications in Computer and Information Science

SP - 161

EP - 174

BT - Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers

A2 - Zhao, Yanchang

A2 - Warwick, Graco

A2 - Stirling, David

A2 - Li, Chang-Tsun

A2 - Koh, Yun Sing

A2 - Islam, Rafiqul

A2 - Islam, Zahidul

PB - Springer-Verlag London Ltd.

ER -

Guo Y, Lin S, Ma X, Bal J, Li CT. Homogeneous feature transfer and heterogeneous location fine-tuning for cross-city property appraisal framework. In Zhao Y, Warwick G, Stirling D, Li C-T, Koh YS, Islam R, Islam Z, editors, Data Mining - 16th Australasian Conference, AusDM 2018, Revised Selected Papers. Springer-Verlag London Ltd. 2019. p. 161-174. (Communications in Computer and Information Science). https://doi.org/10.1007/978-981-13-6661-1_13