Enhancing Regional Airbnb Trend Forecasting Using LLM-Based Embeddings of Accessibility and Human Mobility
Hongju Lee 1, Youngjun Park 2, Jisun An 3, Dongman Lee 2 · 2025
Read Enhancing Regional Airbnb Trend Forecasting Using LLM-Based Embeddings of Accessibility and Human Mobility by Hongju Lee 1, Youngjun Park 2, Jisun An 3, Dongman Lee 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-14248.txt.
Introduction
Hongju Lee, Youngjun Park, Jisun An, and one co-author forecast neighborhood-level Airbnb trends with help from language model embeddings. Short-term rentals strain local housing, so planners need early warning of where revenue and bookings are heading. The framework predicts revenue, reservation days, and reservation counts one to three months ahead using a sliding window.
Instead of modeling single listings, it builds a representation of each region that merges listing features with accessibility and human mobility data. Tables are converted into text prompts, and a language model turns them into regional embeddings for sequence models. On Seoul data the method roughly halved forecast error against conventional baselines.
Catalog record
| CREATOR | Hongju Lee 1, Youngjun Park 2, Jisun An 3, Dongman Lee 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-14248
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Enhancing Regional Airbnb Trend Forecasting Using LLM-Based Embeddings of Accessibility and Human Mobility Hongju Lee 1, Youngjun Park 2, Jisun An 3, Dongman Lee 2 Abstract The expansion of short-term rental platforms, such as Airbnb, has significantly disrupted local housing markets, often leading to increased rental prices and housing affordability issues. Accurately forecasting regional Airbnb market trends can thus offer critical insights for policymakers and urban planners aiming to mitigate these impacts. This study proposes a novel time-series forecasting framework to predict three key Airbnb indicators—Revenue, Reservation Days, and Number of Reservations—at the regional level. Using a sliding-window approach, the model forecasts trends 1 to 3 months ahead. Unlike prior studies that focus on individual listings at fixed time points, our approach constructs regional representations by integrating listing features with external contextual factors such as urban accessibility and human mobility. We convert structured tabular data into prompt-based inputs for a Large Language Model (LLM), producing comprehensive regional embeddings. These embeddings are then fed into advanced time-series models (RNN, LSTM, Transformer) to better capture complex spatio-temporal dynamics. Experiments on Seoul’s Airbnb dataset show that our method reduces both average RMSE and MAE by approximately 48% compared to conventional baselines, including traditional statistical and machine learning models. Our framework not only improves forecasting accuracy but also offers practical insights for detecting oversupplied regions and supporting data-driven urban policy decisions. † † institute: Graduate School of Data Science, KAIST, Daejeon, Republic of Korea, 34141 and School of Computing, KAIST, Daejeon, Republic of Korea, 34141 and Luddy School of Informatics, Computing, and Engineering, Indiana University Bloomington (IUB), USA, 47405 1 Introduction The recent rise of sharing economy platforms such as Airbnb has significantly reshaped local housing markets, triggering both academic research and policy debates on its socioeconomic implications. Empirical studies have shown that regions with concentrated Airbnb activity experience significant increases in rental and housing prices. For instance, in Barcelona, every increase of 54 Airbnb listings corresponded to a 1.9% rise in average rents and a 4.6% rise in housing prices 6 . Likewise, in Los Angeles, the enforcement of the Home Sharing Ordinance (HSO) led to a 50% decrease in Airbnb listings and a subsequent 2% reduction in both rent and housing values 12 . These findings collectively indicate that Airbnb’s proliferation contributes to the displacement of long-term rentals in favor of short-term accommodations, intensifying pressures in the housing market. Thus, developing accurate forecasting methods for regional Airbnb trends has become crucial—not only for academic analysis but also for informing proactive urban policy interventions and regulatory actions aimed at stabilizing housing markets. Previous machine learning studies have mainly focused on predicting Airbnb prices at the individual listing level, often using static features such as amenities, host characteristics, review counts, and proximity to points of interest (POI) 9 ; 11 ; 15 ; 16 . However, these approaches inherently neglect the spatio-temporal dynamics of the Airbnb market. Airbnb listings are typically concentrated near urban cores or tourist attractions, continuously influenced by dynamic factors such as human mobility patterns, transportation infrastructure, and urban accessibility. Moreover, housing market analyses typically require a broader regional context rather than isolated listings. Existing listing-level methods struggle to reflect these regional dynamics, making them insufficient for accurately predicting and responding to shifts in housing market conditions at a neighborhood scale. To overcome these limitations, this study proposes a novel regional-level Airbnb trend forecasting framework based on multivariate time-series analysis. We utilize monthly Airbnb data aggregated at the smallest administrative unit ( dong ) in Seoul, which allows for nuanced regional-level analysis aligned with previous housing market studies 5 ; 6 . Our model targets three core indicators of Airbnb market activity: Revenue , Reservation Days , and Number of Reservations . These indicators, widely recognized in prior research 11 ; 14 ; 15 , directly reflect Airbnb market performance and operational dynamics. Unlike traditional listing-focused approaches, our model explicitly integrates critical external regional context—specifically detailed transportation accessibility (including road networks, transit usage) and fine-grained human mobility patterns (such as domestic and international visitor flows)—to better capture dynamic spatial interactions influencing Airbnb market trends. To effectively encode these heterogeneous and complex contextual factors into a unified representation, we leverage the semantic and contextual understanding capabilities of Large Language Models (LLMs). Specifically, structured regional data are translated into natural-language prompts and processed through LLaMA 3 7 , producing rich embeddings that reflect intricate spatial and temporal dynamics which traditional numeric representations might fail to capture. This study makes several notable contributions to Airbnb forecasting literature: • (1) Regional-level Multivariate Forecasting : Unlike prior studies primarily focusing on predictions at the individual listing level, we explicitly address the regional spatio-temporal dynamics of Airbnb markets using a multivariate time-series framework. Our method effectively captures spatial clustering and temporal dependencies inherent in urban and tourist-centric neighborhoods. • (2) Context-Aware Embeddings via LLM : We propose a novel embedding strategy that leverages external regional context—specifically transportation accessibility and human mobility data—to generate semantically enriched embeddings through LLM-based prompting. This innovative approach substantially improves the representational capability of regional features compared to conventional numeric encoding methods. • (3) Robust and Generalizable Performance Improvements : Empirical evaluations conducted on Seoul’s Airbnb dataset demonstrate that our method consistently achieves an approximately 48% reduction in both RMSE and MAE compared to conventional baseline models, including traditional statistical methods and standard machine learning approaches. These results validate the effectiveness and robustness of our context-aware embedding approach in capturing complex Airbnb market dynamics. Overall, the proposed model offers not only methodological advancements in regional Airbnb trend prediction but also practical implications for identifying oversaturated areas, guiding regulatory interventions, and supporting evidence-based urban policy decisions. 2 Related Work 2.1 Socioeconomic Impacts of Airbnb on Local Housing Markets The rapid expansion of short-term rental platforms like Airbnb has had significant socioeconomic impacts on local housing markets, notably affecting rental prices and housing affordability. Multiple empirical studies have systematically explored these phenomena through spatially explicit regional analyses. Garcia-López et al. 6 utilized spatial econometric methods at Barcelona’s Basic Statistical Area (BSA) scale, demonstrating that Airbnb listing density strongly correlated with increased rents and housing prices, particularly in tourism-centric regions. Similarly, Koster et al. 12 employed a Spatial Regression Discontinuity Design (Spatial RDD) to analyze the policy-induced reduction in Airbnb listings within Los Angeles, revealing a direct causal effect on local housing costs. Franco et al. 5 also reported substantial price escalations linked to Airbnb density at the parish level in Portugal, further highlighting the spatial clustering effects of Airbnb activity on local housing markets. Collectively, these studies consistently reveal that regional Airbnb density is directly associated with increased rental prices and substantial shifts in local housing market dynamics. Notably, they all employ spatially disaggregated, region-level analyses—an approach that closely aligns with the analytical perspective and research motivation of our study. 2.2 Machine Learning Approaches for Airbnb Price Prediction Prior research on Airbnb price prediction has primarily employed machine learning techniques to estimate nightly listing rates using static, listing-level features. Tan et al. 16 integrated multimodal inputs, combining structured metadata, textual descriptions, and image data via deep learning architectures such as BERT and MobileNet, enhancing predictive accuracy. Lektorov et al. 13 utilized user-generated review sentiment alongside