LVRoadNetwork2Vec

A Python-based project for converting Las Vegas traffic sensor locations into graph representations, creating edge lists for network analysis, and visualizing road networks using interactive maps. This project enables graph-based analysis of traffic sensor networks using node2vec embeddings and other graph neural network techniques.

Overview

This project processes traffic sensor data from Las Vegas highways to:

• Clean and normalize sensor location data
• Group sensors by highway/road segments
• Generate edge lists representing sensor connectivity
• Create interactive visualizations using Folium maps
• Enable graph-based machine learning analysis

Features

• Data Cleaning: Remove duplicate sensors and normalize GPS coordinates
• Sensor Grouping: Filter and group sensors by highway identifiers (I-15, US-95, CC-215, I-515)
• Edge List Generation: Two methods for creating graph edges:
  • Minimum Spanning Tree (MST) based on physical distances
  • Sequential adjacency for ordered sensor networks
• Interactive Visualization: Color-coded maps showing sensor locations and connections
• Multi-Highway Support: Visualize multiple highway networks simultaneously

Installation

Prerequisites

• Python 3.7+
• pip package manager

Required Libraries

pip install pandas
pip install folium
pip install geopy
pip install jupyter

Or install all at once:

pip install pandas folium geopy jupyter

Project Structure

LVRoadNetwork2Vec/
├── README.md                           # Project documentation
├── edge_list_maker.ipynb              # Main notebook for edge list creation
├── plotter.ipynb                      # Visualization notebook
├── detector/                          # Traffic sensor data
│   ├── bugatti/                       # BUGATTI system data
│   │   ├── FAST Detectors 2021.09.23 - XIE.xlsx
│   │   └── sensor_description_all.csv
│   └── shredder (UNLV)/              # UNLV detector data
│       ├── detectors2018.csv
│       └── detectors2019.csv
└── graphics/                          # Output visualizations

Data Format

The CSV sensor data files contain the following columns:

Column	Description	Example
0	Index/ID	-
1	Detector/Sensor ID with lane info	I15_NB_SAHARA_lane1
2	-	-
3	Location description	I-15 NB @ Sahara
4	-	-
5	Latitude (compressed format)	3617 → 36.17°
6	Longitude (compressed format)	11513 → -115.13°

Usage

1. Basic Sensor Visualization

Use plotter.ipynb to visualize all sensors:

import folium
import pandas as pd

# Load and process data
data = pd.read_csv('./detector/shredder (UNLV)/detectors2019.csv', header=None)
latitude = data.iloc[:, 5].astype(str).apply(lambda x: float(x[:2] + '.' + x[2:]))
longitude = data.iloc[:, 6].astype(str).apply(lambda x: float(x[:4] + '.' + x[4:]))

# Create map
map = folium.Map(location=[latitude.mean(), longitude.mean()], zoom_start=10)
# Add markers...

2. Create Edge Lists

Use edge_list_maker.ipynb for graph generation:

# Load and clean data
data = clean_dataset(pd.read_csv('./detector/shredder (UNLV)/detectors2018.csv', header=None))

# Filter specific highway
sensor_loc = "I-15 NB"  # Options: "I-15 NB", "I-15 SB", "US-95 NB", "US-95 SB", "CC-215 EB", "CC-215 WB", "I-515 SB", "I-515 NB"
sensor_group = get_sensor_group(data, sensor_loc, ignore='None')

# Generate edge list and visualization
map, edge_list = tsp(sensor_group)
display(map)
print(edge_list)

3. Multi-Highway Visualization

Visualize all highways with color coding:

# Run main() function in edge_list_maker.ipynb
# Automatically generates a map with:
# - Red: CC-215 (East/West)
# - Green: I-15 (North/South)
# - Purple: I-515 (North/South)
# - Yellow: US-95 (North/South)

Core Functions

clean_dataset(df)

Cleans and normalizes the sensor dataset.

Parameters:

• df (DataFrame): Raw sensor data

Returns:

• DataFrame: Cleaned data with:
  • Duplicate sensors removed (based on lat/long)
  • Lane identifiers stripped from sensor names
  • GPS coordinates converted to decimal format

Example:

data = pd.read_csv('./detector/shredder (UNLV)/detectors2018.csv', header=None)
clean_data = clean_dataset(data)

get_sensor_group(data, sensor_loc, ignore=None)

Filters sensors by highway identifier.

Parameters:

• data (DataFrame): Cleaned sensor data
• sensor_loc (str): Highway identifier (e.g., "I-15 NB", "US-95 SB")
• ignore (str, optional): String pattern to exclude from results

Returns:

• DataFrame: Filtered sensor group

Example:

i15_sensors = get_sensor_group(data, "I-15 NB", ignore="-")

calculate_distance(sensor_data)

Calculates pairwise distances between all sensors using GPS coordinates.

Parameters:

• sensor_data (DataFrame): Sensor group data

Returns:

• list[list[float]]: Distance matrix in meters

tsp(sensor_data) - Method 1: MST-based

Creates edge list using Minimum Spanning Tree algorithm based on physical distances.

Parameters:

• sensor_data (DataFrame): Sensor group data

Returns:

• map (folium.Map): Interactive map with sensor connections
• edge_list (DataFrame): Edge list with columns [sensor1, sensor2, cost]

Example Output:

  sensor1         sensor2         cost
  I15_NB_SAHARA  I15_NB_CHARLESTON  1247.3
  I15_NB_OAKEY   I15_NB_SAHARA      891.2

tsp(sensor_data) - Method 2: Sequential Adjacency

Creates 0-1 edge list where consecutive sensors are connected.

Returns:

• map (folium.Map): Interactive map
• edge_list (DataFrame): Edge list with columns [sensor1, sensor2, weight]
  • weight=1: Adjacent sensors
  • weight=0: Non-adjacent sensors

Example Output:

  sensor1  sensor2  weight
  1        2        1
  1        3        0
  2        3        1

Highway Identifiers

The following highway segments are available in the dataset:

• I-15: Interstate 15 (Northbound/Southbound)
• US-95: US Route 95 (Northbound/Southbound)
• CC-215: Clark County 215 Beltway (Eastbound/Westbound)
• I-515: Interstate 515 (Northbound/Southbound)

Applications

This project enables various graph-based analyses:

1. Node2Vec Embeddings: Generate vector representations of sensors for ML models
2. Traffic Flow Analysis: Study traffic patterns across connected sensors
3. Network Coverage: Analyze sensor distribution and identify coverage gaps
4. Graph Neural Networks: Train GNNs on the sensor network topology
5. Anomaly Detection: Identify unusual traffic patterns using graph structure

Output Files

• Interactive Maps: HTML files saved using map.save('output.html')
• Edge Lists: CSV files with sensor connectivity
• Visualizations: Stored in graphics/ directory

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues for bugs and feature requests.

Data Sources

• UNLV Shredder Data: 2018-2019 traffic detector data
• BUGATTI System: FAST Detectors data (2021)

License

MIT License

Citation

If you use this project in your research, please cite:

T. Bin Zahid and B. T. Morris, "Benchmarking/Limitations of Traffic Prediction with Noisy Field Measurements," 2024 IEEE International Conference on Vehicular Electronics and Safety (ICVES), Ahmedabad, India, 2024, pp. 1-6, doi: 10.1109/ICVES61986.2024.10928136. keywords: {Training;Vehicular and wireless technologies;Accuracy;Roads;Urban planning;Predictive models;Transformers;Data models;Robustness;Noise measurement},

T. b. Zahid and B. Morris, "Using Deep Traffic Prediction for EMFAC Emission Estimation and Visualization," 2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), Edmonton, AB, Canada, 2024, pp. 2488-2493, doi: 10.1109/ITSC58415.2024.10919675. keywords: {Solid modeling;Accuracy;Decision making;Transportation;Estimation;Data visualization;Predictive models;Transformers;Data models;Environmental factors},

---

Last Updated: January 2026