Self-supervised Urban Parcel Representation Considering Multi-semantic Feature Fusion of POIs

About PMSF-PRPrecise parcel representation is essential for urban planning, resource allocation, and socioeconomic forecasting. Relying on single semantic features and annotated data limits both the quality and generalizability of parcel representations in most existing POI-based methods. This study proposes PMSF-PR, a self-supervised parcel representation learning method that incorporates multi-semantic POI feature fusion. PMSF-PR integrates POI geographic semantics, hierarchical category semantics, and textual features derived from a large language model (LLM) through a multi-semantic feature fusion mechanism that captures geographic heterogeneity. Trained via self-supervised contrastive learning without using labeled data, PMSF-PR enables large-scale parcel representation across China. In the parcel retrieval task, we built an efficient vector database and accurately identified parcels with similar functional attributes. Quantitative evaluations on three downstream tasks show that PMSF-PR outperforms baseline methods, improving land use classification accuracy by 2.63% to 8.91%, and increasing Pearson’s R by 28.37% to 46.31% for GDP estimation and 4.00% to 5.58% for population density estimation. Ablation experiments further validate the effectiveness of geographic heterogeneity modeling and the integration of multiple semantic features. This study provides a scalable and label-free parcel representation method for urban spatial analysis, with promising applications in smart city development and socioeconomic forecasting.Please see “PMSF-PR Project Document.pdf” for code implementation details. For more information on how to reproduce the results presented in the paper, refer to this document.

Authors

Publication Details

Journal
Figshare
Published
2026-11-21
DOI
https://doi.org/10.6084/m9.figshare.29320316
Primary Topic
Geographic Information Systems Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Self-supervised Urban Parcel Representation Considering Multi-semantic Feature Fusion of POIs

Yao Yao
Figshare
Geographic Information Systems Studies
article

Self-supervised Urban Parcel Representation Considering Multi-semantic Feature Fusion of POIs

Yao Yao
article en

Abstract

About PMSF-PRPrecise parcel representation is essential for urban planning, resource allocation, and socioeconomic forecasting. Relying on single semantic features and annotated data limits both the quality and generalizability of parcel representations in most existing POI-based methods. This study proposes PMSF-PR, a self-supervised parcel representation learning method that incorporates multi-semantic POI feature fusion. PMSF-PR integrates POI geographic semantics, hierarchical category semantics, and textual features derived from a large language model (LLM) through a multi-semantic feature fusion mechanism that captures geographic heterogeneity. Trained via self-supervised contrastive learning without using labeled data, PMSF-PR enables large-scale parcel representation across China. In the parcel retrieval task, we built an efficient vector database and accurately identified parcels with similar functional attributes. Quantitative evaluations on three downstream tasks show that PMSF-PR outperforms baseline methods, improving land use classification accuracy by 2.63% to 8.91%, and increasing Pearson’s R by 28.37% to 46.31% for GDP estimation and 4.00% to 5.58% for population density estimation. Ablation experiments further validate the effectiveness of geographic heterogeneity modeling and the integration of multiple semantic features. This study provides a scalable and label-free parcel representation method for urban spatial analysis, with promising applications in smart city development and socioeconomic forecasting.Please see “PMSF-PR Project Document.pdf” for code implementation details. For more information on how to reproduce the results presented in the paper, refer to this document.

Figshare
Geographic Information Systems Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
Self-supervised Urban Parcel Representation Considering Multi-semantic Feature Fusion of POIs — Yao Yao · Figshare (2026) | TGRS Research Map | TGRS