LatamBodyguard is a web application developed during UAB The Hack that helps users find the safest and fastest routes between two locations in Mexico City using crime data analysis and weighted graph algorithms.

This solution was created for PayRetailers' challenge focused on solving real-world problems in South America. LatamBodyguard combines geospatial analysis with AI to provide route recommendations that prioritize both safety and efficiency.

• Safety-Optimized Routes: Generates paths that avoid high-crime areas while maintaining efficiency
• AI-Powered Chatbot: Uses Large Language Models to explain routing decisions with transparency
• Crime Data Integration: Analyzes Mexico City crime statistics to inform route calculations
• Cluster Analysis: Employs buffer-based clustering to distinguish between types of urban areas

1. Crime Data Collection: We aggregated and processed crime incident data from Mexico City, categorizing by type, severity, and location.
2. Geospatial Mapping: The application creates a heat map of criminal activity across the city, identifying high-risk zones.
3. Buffer Zones: We implemented a buffer system around high-crime areas, with varying levels of "danger radius" based on crime type and frequency.

1. Graph Construction: The city map is converted into a weighted graph where:
   • Nodes represent intersections and key points
   • Edges represent streets and paths
   • Edge weights incorporate both distance and a calculated safety score
2. Safety Weighting: Each path segment is assigned a safety score derived from:
   • Proximity to high-crime areas
   • Time of day (adjusts risk factors based on historical crime patterns)
   • Crime type prevalence in the area
3. Path Calculation: We use a modified A* algorithm that optimizes for both shortest distance and maximum safety, allowing the system to find routes that may be slightly longer but significantly safer.

1. Query Input: Users enter starting and destination points
2. Route Generation: The system calculates and displays the optimal safe route
3. AI Explanation: The integrated LLM chatbot provides natural language explanations of:
   • Why certain areas were avoided
   • Trade-offs made between speed and safety
   • Specific safety concerns along the route

• Backend: Python-based geospatial processing using GeoPandas and NetworkX for graph operations
• AI Component: Integration of LLM APIs for natural language processing and response generation
• Frontend: Interactive map visualization using Leaflet.js with custom safety overlay layers
• Data Pipeline: Automated processing of crime statistics with regular update capability

• Sergi Flores
• Jiahui Chen
• Weihao Lin
• Clàudia Gallego

• Geospatial Analysis
• Weighted Graph Algorithms
• Large Language Models
• Web Development

• Data Integration: Harmonizing different crime data sources and formats
• Algorithm Optimization: Balancing computational efficiency with route quality
• UX Design: Creating an intuitive interface that clearly communicates safety information without causing alarm

LatamBodyguard demonstrates the practical application of AI and data science to address real safety concerns in urban environments, offering a solution that could potentially improve daily life for people in high-crime areas of Latin America.