IndustryScopeKG is the first and largest open-source, multimodal large-scale knowledge graph dataset. It integrates geospatial data such as street view images, points of interest, and crowd activities, along with socio-economic data including company information, real estate prices, and population statistics. The dataset contains 2,232,037 entities and 51,684,939 triples.
The IndustryScopeGPT framework was developed to enhance the planning, action, and reasoning capabilities of large language models (LLMs) in industrial park planning and operations. This framework represents the first integration of large language models with spatial computing and dynamic reasoning on graph databases that incorporate external geospatial data, marking the first exploration of such applications in industrial park planning and operations.
2024/10/28We’re thrilled to share that our work, Decoding Urban Industrial Complexity: Enhancing Knowledge-Driven Insights via IndustryScopeGPT, has been published by the ACM MM 2024 conference. We are profoundly grateful for the contributions of all the contributors to this work！2024/07/23We are honored to share and present our work at the MIT Media Lab-City Science Lab Community Meeting. For more details, please visit the City Science Lab website: Design and Data-Driven Hybrid Community Building.
The raw data is available for download from Kaggle or Baidu Netdisk.
industryscopekg.dump
This file contains the standalone Knowledge Graph data, designed for users who want to work directly with the raw data or already have a Neo4j setup and wish to directly import the dataset.
industryscopekgdocker.zip
If you downloaded the file from Kaggle as industryscopekgdocker.sip, please rename the file extension from .sip to .zip and unzip the contents.
If you obtained the file from Baidu Netdisk, it will already be named industryscopekgdocker.zip, and no further renaming is required. Simply unzip the contents to proceed.
After unzipping the .zip file, you will get the industryscopekgneo4j.tar file.
For seamless integration, we recommend using Neo4j to analyze and interact with the knowledge graph.
We provide a Neo4j Docker Image to simplify the process of setting up and working with the IndustryScopeKG dataset.
Ensure you have Docker installed on your system. For installation instructions, visit Docker's official website.
Use the following command to load the industryscopekgneo4j.tar into your local Docker environment:
docker load -i industryscopekgneo4j.tarAfter the image is loaded, Docker will display the name and tag of the image, typically neo4jv_backup:latest.
Run the following command to start a Neo4j container with the dataset preloaded:
docker run \
-itd \
--name industryscope \
-p 7474:7474 -p 7687:7687 \
neo4jv_backup:latestDefault credentials:
- Username:
neo4j - Password:
bitnami1
Note: We use bitnami/neo4j:5.13.0 as our base image, and we compiled spatial plugin manually with maven which already exists in our docker image provided.
- Open your browser and navigate to: http://localhost:7474.
- Login with the credentials (
neoo4j/bitnami1).
Here are some useful Cypher queries you can play with:
Question 1: 静安区的所有产业园区都有哪些？
MATCH (p:park)
WHERE p.所属区县 = "静安区"
RETURN p.名称 AS 园区名称
Question 2: 上海张江高科技园内的网格根据零售服务POI数量的排序是什么？
MATCH (p:park {{名称: "上海张江高科技园区"}})-[:IS_IN]-(f:fishnet)
RETURN p.fishnet AS FishnetID, f.零售服务POI数量 AS RetailServicesCount
ORDER BY RetailServicesCount DESC
Question 3: 朕天总部经济工业园里有没有便利店？
MATCH (poi:poi_retail)-[:IS_IN]->(:park {{名称: '朕天总部经济工业园'}})
WHERE poi.类别 CONTAINS '便利店' OR poi.类别 CONTAINS '零售' OR poi.类别 CONTAINS '便民商店'
RETURN COUNT(poi) AS 便利店数量
Question 4: 张江高新区杨浦园里有几家华润？
MATCH (poi)-[:IS_IN]->(:park {{名称: '张江高新区杨浦园'}})
WHERE poi.名称 CONTAINS '华润'
RETURN COUNT(poi) AS 华润数量
Question 5: 张江高新区杨浦园里的华润名称是什么？
MATCH (poi)-[:IS_IN]->(:park{{名称: '张江高新区杨浦园'}})
WHERE poi.名称 CONTAINS '华润'
RETURN poi.名称 AS 华润名称
Table 1: The Statistics of Entities in IndustryScopeKG
| Basic Statistics | Industrial Park | Grid | Grid Dominant Function | POI | Enterprise | Total |
|---|---|---|---|---|---|---|
| Count | 264 | 128,866 | 15 | 112,931 | 1,058,656 | 1,300,732 |
| (Leading) Industries | Primary | Secondary | Tertiary | Scope of Operations | Planned | |
| Industrial Park | 202 | 258 | 261 | 261 | 70 | 1,052 |
| Grid | 1,142 | 6,270 | 10,281 | 20,246 | / | 37,939 |
| Enterprise | 18 | 90 | 392 | 891,814 | / | 892,314 |
Table 2: The Statistics of Triples in IndustryScopeKG
| Relation | Head & Tail Entity | Triple Records |
|---|---|---|
| Locate in | (POI, Grid), (Enterprise, Grid), (POI, Industrial Park), (Enterprise, Industrial Park), (Grid, Industrial Park) | 2,516,160 |
| Adjacent to | (Grid, Grid), (Industrial Park, Industrial Park) | 488,401 |
| Similar to | (Industrial Park, Industrial Park) | 3,765 |
| Related to | (Industrial Park, Industrial Park) | 10,687 |
| Has | E.g., (Industrial Park, Planned Industries), (Grid, Leading Scope of Operations), (Grid, Dominant Functions) | 4,252,341 |
| Attribution | (Industrial Park, Value) with 111 attributions, (Grid, Value) with 82 attributions, (POI, Value) including 15 attributions, (Enterprise, Value) with 36 attributions | 44,413,585 |
The dataset can be used standalone or integrated into your projects. It works particularly well with graph-based reasoning frameworks. For LLM-based planning and operations, the dataset complements the IndustryScopeGPT framework, providing a dynamic and insightful environment for urban and industrial research.
IndustryscopeGPT scripts will be available soon under the folder scripts.
If our work sparks your interest or helps your research, a little cite would mean a lot!
@inproceedings{10.1145/3664647.3681705,
author = {Wang, Siqi and Liang, Chao and Gao, Yunfan and Liu, Yang and Li, Jing and Wang, Haofen},
title = {Decoding Urban Industrial Complexity: Enhancing Knowledge-Driven Insights via IndustryScopeGPT},
year = {2024},
isbn = {9798400706868},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3664647.3681705},
doi = {10.1145/3664647.3681705},
abstract = {Industrial parks are critical to urban economic growth. Yet, their development often encounters challenges stemming from imbalances between industrial requirements and urban services, underscoring the need for strategic planning and operations. This paper introduces IndustryScopeKG, a pioneering large-scale multi-modal, multi-level industrial park knowledge graph, which integrates diverse urban data including street views, corporate, socio-economic, and geospatial information, capturing the complex relationships and semantics within industrial parks. Alongside this, we present the IndustryScopeGPT framework, which leverages Large Language Models (LLMs) with Monte Carlo Tree Search to enhance tool-augmented reasoning and decision-making in Industrial Park Planning and Operation (IPPO). Our work significantly improves site recommendation and functional planning, demonstrating the potential of combining LLMs with structured datasets to advance industrial park management. This approach sets a new benchmark for intelligent IPPO research and lays a robust foundation for advancing urban industrial development. The dataset and related code are available at https://github.com/Tongji-KGLLM/IndustryScope.},
booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia},
pages = {4757–4765},
numpages = {9},
keywords = {industrial park planning and operation, large language model agent, urban design and planning, urban knowledge graph},
location = {Melbourne VIC, Australia},
series = {MM '24}
}