Research on the Sustainable Development of Urban Housing Price Based on Transport Accessibility: A Case Study of Xi’an, China

Chao Xue, Yongfeng Ju, Shuguang Li, Qilong ZhouSustainability 2020 · DOI

TL;DR

This paper introduces a novel approach to predicting urban housing prices by incorporating bus and metro accessibility, calculated using space syntax and OD cost matrix methods, into a Random Forest (RF) model. The study demonstrates that these transport accessibility metrics significantly improve prediction accuracy compared to traditional hedonic price models and RF models without these features, highlighting the crucial role of public transit in real estate valuation.

What Problem Does This Paper Address?

Accurately predicting urban housing prices is a complex challenge due to the multitude of influencing factors. While traditional models often focus on internal property attributes and general location factors, they tend to overlook the nuanced impact of transportation accessibility. This is particularly relevant in the context of sustainable urban development, where efficient public transport is a key component. Existing methods for quantifying transport accessibility, such as simple distance metrics, may not fully capture the real-world convenience and connectivity offered by public transit networks. This paper aims to bridge this gap by developing and evaluating a more sophisticated method for measuring bus and metro accessibility and integrating it into a robust predictive model.

Methods

The study proposes a novel method for quantifying transport accessibility. For bus accessibility, they utilize an OD (origin-destination) cost matrix approach, calculating the minimum impedance (shortest path distance) between bus stops, weighted by the number of bus lines serving each stop. For metro accessibility, they employ space syntax theory, specifically the "degree of integration" metric, to assess how well each metro station is connected within the overall metro network.

These accessibility metrics are then integrated into two primary modeling frameworks:

  1. Traditional Hedonic Price Model (HPM): A standard econometric approach for real estate valuation.
  2. Random Forest (RF) Model: A machine learning algorithm known for its ability to handle complex, non-linear relationships and high-dimensional data.

Four distinct models are constructed and compared:

  • HPM without transport accessibility (M1)
  • HPM with transport accessibility (M2)
  • RF without transport accessibility (M3)
  • RF with transport accessibility (M4)

The models are evaluated using R² (coefficient of determination) and RMSE (root mean square error) on a testing set, with a 7:3 train-test split.

Datasets

The study utilizes data from the main urban area of Xi'an, China.

  • House property data: Collected from ANJUKE, a Chinese real estate information website (June 2019).
  • Urban transportation data: Including road networks, bus routes, and metro lines, obtained via the GAODE app API (May 2019).

The paper does not introduce new benchmark datasets but rather uses publicly available data from a specific city for its case study.

Key Insights and Ideas

  • Transport accessibility is a significant driver of housing prices: The study strongly suggests that how easily residents can access public transport (buses and metros) has a substantial impact on property values, often rivaling or exceeding the importance of some traditional property features.
  • Space syntax and OD cost matrix offer superior accessibility metrics: These methods provide a more nuanced and realistic representation of transport convenience compared to simple distance calculations, better reflecting the actual urban transit system's structure and connectivity.
  • Random Forest excels in this complex prediction task: The RF model, especially when augmented with transport accessibility features, significantly outperforms the traditional HPM, demonstrating its capability to capture intricate relationships in real estate data.
  • Sustainable urban planning benefits from accurate accessibility metrics: By providing data-driven insights into how transport infrastructure influences housing prices, this research can inform urban spatial planning decisions, promoting balanced regional development and sustainable growth.

Key Results

  • Model M4 (RF with transport accessibility) achieved the highest prediction accuracy: It yielded an R² of 0.840 and an RMSE of 2142 on the testing set.
  • M1 (Traditional HPM without transport) performed poorly: It had an R² of 0.219 and an RMSE of 4801, indicating low accuracy.
  • Adding transport accessibility improved HPM: M2 (HPM with transport) showed a slight improvement over M1 (R²=0.227, RMSE=4735), but still lagged far behind RF models.
  • Transport features are highly important: In the RF model (M4), bus accessibility and metro accessibility were among the top three most important features, alongside housing area.
  • Real-time performance is adequate: All models, including the more complex RF, completed predictions in under one second, ensuring practical real-time application.

Limitations and Future Work

The authors acknowledge several limitations:

  • Omission of external amenities: Factors like hospitals, shopping malls, and schools were not included, which could influence housing prices.
  • Limited model comparison: Only HPM and RF were compared; exploring other advanced models could yield further insights.
  • Geographic scope: The study is limited to Xi'an, China. Expanding to cities with diverse spatial characteristics would enhance generalizability.

Future work suggestions include:

  • More refined accessibility calculations: Particularly for bus accessibility, obtaining actual route data (instead of relying solely on shortest paths) would improve accuracy.
  • Inclusion of more influencing factors: Incorporating variables like noise levels, crime rates, security, and green areas could further enhance model performance.
  • Application to sustainable urban development: Exploring how these models can be used to analyze regional development imbalances and promote sustainable urban growth.

Summary updated

Abstract

The development of a real estate economy is beneficial to urban stability. A method of real estate price prediction based on transport accessibility is proposed. The method adds bus accessibility and metro accessibility into the model, which has higher prediction accuracy than the traditional model. Firstly, bus accessibility and metro accessibility are calculated according to the space syntax theory. Then, four models, the traditional hedonic price model (HPM) with transport accessibility, the traditional hedonic price model without transport accessibility, the random forest (RF) model with transport accessibility, and the random forest model without transport accessibility, are introduced. Finally, the four models are compared and analyzed in terms of precision and importance of index contributions. Taking Xi ’an, China, as an example, the experimental results show that the transport accessibility calculated based on space syntax can accurately represent the transport convenience in an urban space structure. Furthermore, it has a great influence on the contribution of indexes in the model. With the introduction of bus accessibility and metro accessibility, the accuracy of the real estate price prediction model is greatly improved.
BibTeX
@inproceedings{xue2020research,
  author = {Chao Xue and Yongfeng Ju and Shuguang Li and Qilong Zhou},
  title = {{Research on the Sustainable Development of Urban Housing Price Based on Transport Accessibility: A Case Study of Xi’an, China}},
  booktitle = {Sustainability},
  year = {2020},
  doi = {10.3390/su12041497},
}
Figure