____ _ ____ _ _
 / ___|_ __ __ _ _ __ | |__ | _ \ __ _| |_ __ _ ___ ___| |_ ___
 | | _| '__/ _` | '_ \| '_ \ | | | |/ _` | __/ _` / __|/ _ \ __/ __|
 | |_| | | | (_| | |_) | | | | | |_| | (_| | || (_| \__ \ __/ |_\__ \
 \____|_| \__,_| .__/|_| |_| |____/ \__,_|\__\__,_|___/\___|\__|___/
 |_|

License: MIT Python 3.8+ Format: CSV

A collection of scripts to download and convert popular graph datasets into a unified CSV format for benchmarking graph databases and algorithms.

• 🎯 Unified Format: All datasets converted to consistent nodes.csv + edges.csv format
• 📊 30+ Datasets: From small test graphs to billion-edge networks
• 🚀 Easy to Use: Simple make commands to download and convert
• 🔄 Multiple Sources: Support for MTX, OGB, Yelp, and more
• 📈 Progress Tracking: Built-in progress bars for large downloads
• 💾 Smart Caching: Skip downloads if files already exist

# Build all datasets
make
# Build one or more specific datasets directly
make ak2010
make ak2010 belgium_osm soc-LiveJournal1
# Other targets work the same way
make fetch ak2010 belgium_osm # download only, no conversion
make clean ak2010 # clean a specific dataset

All datasets are converted to a consistent, simple format for easy integration:

node_id
0
1
2
...

src,dst
0,1
0,2
1,3
...

src,dst,weight
0,1,0.5
0,2,1.0
...

• ✅ Contiguous 0-based node IDs: All node IDs are remapped to a contiguous sequence starting from 0
• ✅ UTF-8 encoded: Universal compatibility
• ✅ Header row: Column names in first line
• ✅ Comma-delimited: Standard CSV format
• ✅ Optional properties: Extensible with additional columns (e.g., type, label)

Note: Node IDs are always remapped to a contiguous 0-based sequence [0, 1, 2, ..., N-1], regardless of the original IDs in the source dataset. This ensures consistent and efficient indexing across all datasets.

• soc-LiveJournal1 - LiveJournal social network
• soc-orkut - Orkut social network
• soc-twitter-2010 - Twitter follower network
• soc-sinaweibo - Sina Weibo social network

• cit-Patents - Patent citation network
• coAuthorsDBLP - DBLP co-authorship network

• roadNet-CA - California road network
• road_usa - USA road network
• road_central - Central USA road network
• belgium_osm - Belgium OpenStreetMap
• germany_osm - Germany OpenStreetMap
• europe_osm - Europe OpenStreetMap
• asia_osm - Asia OpenStreetMap
• osm-road-networks - Any city via OpenStreetMap (osmnx), with road attributes (lat/lon, speed, travel time, highway type, etc.)

• uk-2002 - UK web graph (2002)
• uk-2005 - UK web graph (2005)
• arabic-2005 - Arabic web graph
• indochina-2004 - Indochina web graph
• webbase-1M - WebBase crawl (1M nodes)
• webbase-2001 - WebBase crawl (2001)

• delaunay_n13 - Delaunay triangulation (2^13 nodes)
• delaunay_n21 - Delaunay triangulation (2^21 nodes)
• delaunay_n24 - Delaunay triangulation (2^24 nodes)
• kron_g500-logn21 - Kronecker graph

• ogbn-products - Amazon product co-purchase network (OGB)
• yelp - Yelp user-business review network (bipartite)
• imdb - IMDB title-person bipartite network (movies, shows, cast & crew)
• movielens-small - MovieLens small rating dataset (~100K ratings)
• movielens - MovieLens full rating dataset (~33M ratings)
• ldbc-snb - LDBC Social Network Benchmark

• hollywood-2009 - Hollywood actor collaboration network
• ak2010 - Autonomous systems graph
• geolocation - Geolocation network

import pandas as pd
# Load graph
nodes = pd.read_csv('nodes.csv')
edges = pd.read_csv('edges.csv')
print(f"Nodes: {len(nodes)}, Edges: {len(edges)}")
# For typed nodes (e.g., Yelp)
if 'type' in nodes.columns:
 print(nodes['type'].value_counts())

#include <fstream>
#include <sstream>
#include <vector>
struct Edge { int src, dst; };
std::vector<Edge> read_edges(const std::string& filename) {
 std::vector<Edge> edges;
 std::ifstream file(filename);
 std::string line;
 std::getline(file, line); // Skip header
 while (std::getline(file, line)) {
 std::istringstream iss(line);
 Edge e;
 char comma;
 iss >> e.src >> comma >> e.dst;
 edges.push_back(e);
 }
 return edges;
}

• mtx2csv.py - Convert Matrix Market (.mtx) to CSV
• ogbn-products/ - OGB dataset converter
• yelp/ - Yelp dataset converter
• osm-road-networks/ - OSM road network downloader (osmnx)
• preview_graph.py - Preview graph statistics

Download any city's road network using osmnx:

# Default city (Pasadena, CA)
make osm-road-networks
# Custom city
make -C osm-road-networks PLACE="Beijing, China"
make -C osm-road-networks PLACE="Tokyo, Japan"

Output per city (in a subdirectory named after the place):

osm-road-networks/
 pasadena_california_usa/
 nodes.csv # node_id, lat, lon
 edges.csv # src, dst, length, speed_kph, travel_time, name, highway, oneway, maxspeed, lanes
 beijing_china/
 nodes.csv
 edges.csv

Some datasets include additional node/edge properties:

Yelp (bipartite graph):

node_id,type,stars,review_count
0,business,4.0,12
150346,user,3.72,15
...

Extensible format - add columns as needed:

src,dst,weight,timestamp,label
0,1,0.5,1609459200,friend

• ✅ MTX (Matrix Market)
• ✅ OGB (Open Graph Benchmark)
• ✅ Yelp JSON
• ✅ SNAP format
• ✅ Custom formats

Contributions welcome! To add a new dataset:

1. Create a subdirectory with dataset name
2. Add a Makefile with download/conversion rules
3. Ensure output follows the unified CSV format
4. Update this README

MIT License - see individual dataset sources for their respective licenses.

• SNAP
• Network Repository
• SuiteSparse Matrix Collection
• OGB
• LDBC

Note: Dataset sizes range from thousands to billions of edges. Check individual dataset directories for specific statistics and download requirements.