Road Traffic Accidents Analysis In Mexico City Through Crowdsourcing Data And Data Mining Techniques

Perez, Gabriela V. Angeles; Lopez, Jose Castillejos; Cabello, Araceli L. Reyes; Grajales, Emilio Bravo; Espinosa, Adriana Perez; Fabian, Jose L. Quiroz · 2018 · OpenAlex-citations

DOI: 10.5281/zenodo.1340531

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of analyzing road traffic accidents in Mexico City (CDMX), a major contributor to urban congestion and mortality. Traditional accident analysis methods are often costly, time-consuming, and lack real-time relevance. The authors propose a low-cost, efficient alternative using crowdsourced data from the Waze navigation app, leveraging data mining techniques to identify high-risk zones, roads, and time periods. The research aims to provide current, actionable insights into accident patterns to support traffic management and safety initiatives. The methodology follows the Knowledge Discovery in Databases (KDD) process. Data was collected via a Python script that queried the Waze website every ten minutes, retrieving accident reports in JSON format for the year 2016. The dataset was filtered to include only reports classified as "ACCIDENT" within CDMX boundaries. To consolidate multiple reports of the same incident, the authors applied Road Safety Theory criteria: reports were grouped if they shared the same subtype, occurred within 150 meters (calculated using the Haversine equation), and fell within a 20-minute window. This process generated a set of "representative" records for each accident. June 2016 was selected for detailed analysis as it had the highest number of accident reports. Data mining was performed using the Weka tool and R scripts. The Expectation Maximization (EM) algorithm determined the optimal number of clusters, while the K-means algorithm grouped the data. Results were visualized using the Geographic Information System QGIS. The analysis focused on 9,946 accident representatives from June 2016. The EM algorithm suggested 17 clusters, but those representing less than 3% of the data were excluded, resulting in 10 significant clusters for the K-means analysis. The results identified specific spatial and temporal patterns. The borough of Miguel Hidalgo contained the largest number of accident representatives. The most common accident type was "ACCIDENT MINOR." Specific clusters revealed distinct hotspots; for instance, Cluster 5, the largest group with 1,706 instances, was located in Miguel Hidalgo, involving minor accidents primarily on July 7th at 12:00. Other clusters identified significant activity in boroughs such as Venustiano Carranza, Coyoacán, Gustavo A. Madero, Álvaro Obregón, Benito Juárez, Iztapalapa, and Cuauhtémoc. The data indicated that accidents were not uniformly distributed but concentrated in specific geographic coordinates and time intervals, with peak hours varying by cluster (e.g., 12:00, 20:00, and 21:00). The study demonstrates that crowdsourced data from social navigation platforms like Waze, combined with clustering algorithms like K-means and EM, can effectively identify traffic accident patterns in urban environments. This approach offers a viable, low-cost alternative to traditional data collection methods, enabling real-time monitoring and analysis. By pinpointing specific high-risk areas and times, the findings can inform targeted traffic safety interventions and urban planning decisions in Mexico City and potentially other metropolitan areas.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash