Options
Modelling housing market dynamics through social media: Mixed artificial intelligence and spatial approaches
Author(s)
Date Issued
2025
Publisher
Hong Kong: Hong Kong Shue Yan University
Description
320 pages
Type
Thesis
Programme
Doctor of Philosophy in Economics
Abstract
As China’s real estate market faces a substantial decline, understanding the patterns of housing prices becomes increasingly crucial. This study investigates the connection between housing price trends and public sentiment by combining qualitative and quantitative methods, aiming to understand how sentiment influences market behaviour.
Using data from China’s top 50 cities and social media comments on Weibo, sentiment scores are calculated with SnowNLP. We then perform annual average sentiment clustering using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). Additionally, we create a Sentiment-to-Housing Price Index to measure the gap between market sentiment and housing prices, and employ K-means clustering to display clusters. Moreover, the Maximal Information Coefficient (MIC) is used to evaluate the correlation between sentiment and housing prices. By presenting the results of MIC, Granger Causality, and Causal Forest Analysis, this study provides comprehensive insights into China's real estate market in recent years.
A Spatial Error Model (SEM) is used to explore potential factors influencing housing market sentiment. The results show that higher housing prices and income inequality correlate with more negative sentiment. Conversely, green coverage, retail sales, and housing supply positively affect sentiment. A significant spatial error component (λ) suggests that unobservable spatially correlated factors, such as intercity migration, cross-regional investment, and regional disparities, play a crucial role in shaping sentiment.
Policy implications include enhancing sentiment monitoring systems, promoting urban greenery, and addressing inequality and emotional spillovers. As housing remains a key source of household wealth in China, maintaining stable and positive public sentiment is vital for sustainable real estate development.
This research contributes to the literature by bridging gaps between qualitative and quantitative sentiment analysis in the real estate sector. It demonstrates the value of social media as a gauge of public opinion and highlights the potential of artificial intelligence and data mining in forecasting market dynamics. The results suggest that government efforts to monitor and respond to sentiment may help restore confidence and stabilise the real estate market.
Using data from China’s top 50 cities and social media comments on Weibo, sentiment scores are calculated with SnowNLP. We then perform annual average sentiment clustering using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). Additionally, we create a Sentiment-to-Housing Price Index to measure the gap between market sentiment and housing prices, and employ K-means clustering to display clusters. Moreover, the Maximal Information Coefficient (MIC) is used to evaluate the correlation between sentiment and housing prices. By presenting the results of MIC, Granger Causality, and Causal Forest Analysis, this study provides comprehensive insights into China's real estate market in recent years.
A Spatial Error Model (SEM) is used to explore potential factors influencing housing market sentiment. The results show that higher housing prices and income inequality correlate with more negative sentiment. Conversely, green coverage, retail sales, and housing supply positively affect sentiment. A significant spatial error component (λ) suggests that unobservable spatially correlated factors, such as intercity migration, cross-regional investment, and regional disparities, play a crucial role in shaping sentiment.
Policy implications include enhancing sentiment monitoring systems, promoting urban greenery, and addressing inequality and emotional spillovers. As housing remains a key source of household wealth in China, maintaining stable and positive public sentiment is vital for sustainable real estate development.
This research contributes to the literature by bridging gaps between qualitative and quantitative sentiment analysis in the real estate sector. It demonstrates the value of social media as a gauge of public opinion and highlights the potential of artificial intelligence and data mining in forecasting market dynamics. The results suggest that government efforts to monitor and respond to sentiment may help restore confidence and stabilise the real estate market.
File(s)
Name
229702D CHEN Siming.pdf
Description
Embargo period 2025-2028
Size
4.33 MB
Format
Adobe PDF
Checksum
(not present)
(MD5):5cd796e2992fe7e9c8d6d9b05ab553ad
Availability at HKSYU Library
