CB-SMoT

The CB-SMoT (Clustering-Based Approach for Discovering Interesting Places in a Single Trajectory)

Identificação semântica de regiões de congestionamento no transporte público da cidade de São Paulo, utilizando API Olho Vivo

Slides: CB-SMoT Project Presentation

Data Analysis and Reporting Enhancements

This project has been enhanced with new features focused on persistent storage of detected low-speed segments, streamlined data processing, and comprehensive statistical reporting. These additions significantly improve the project's data science capabilities by enabling:

• Structured Analysis: Storing detected segments in a database allows for more organized and repeatable analysis.
• Feature Engineering: The attributes of detected segments (e.g., duration, average speed, geometry) can serve as features for further machine learning models or deeper data mining.
• Quantitative Insights: Statistical reports provide measurable insights into traffic patterns, congestion hotspots, and operational characteristics of public transport.
• Historical Tracking: Persisting segments allows for tracking changes and trends over time if data is processed periodically.

New Django Model: DetectedSegment

This model is designed to store the low-speed or "stop" segments identified from vehicle trajectories by the CB-SMoT algorithm.

Key Fields:

• vehicle: A foreign key to the Vehicle model, indicating which vehicle generated the segment.
• route: A foreign key to the Route model, indicating the route the vehicle was on (if available).
• start_time: DateTimeField indicating the beginning of the segment.
• end_time: DateTimeField indicating the end of the segment.
• duration_seconds: FloatField storing the segment's duration in seconds.
• distance_meters: FloatField storing the distance covered within the segment in meters.
• num_points: IntegerField for the number of trajectory points that constitute the segment.
• average_speed_mps: FloatField for the calculated average speed within the segment in meters per second.
• trajectory_points: A ManyToManyField linking to the Trajectory model, storing the actual points that make up this segment.
• segment_geometry: A LineStringField (GIS) storing the geographical path of the segment.

New Management Commands

Two new Django management commands have been introduced to facilitate the processing and analysis of trajectory data.

1. process_trajectories

This command processes raw trajectory data (assumed to be populated in the Trajectory model) using the underlying CB-SMoT algorithm. It identifies low-speed segments and saves them as DetectedSegment instances in the database.

Key Command-Line Arguments:

• --vehicle_ids <id1> <id2> ...: Process trajectories only for the specified vehicle IDs.
• --route_ids <id1> <id2> ...: Process trajectories for all vehicles associated with the specified route IDs. (Ignored if --vehicle_ids is used).
• --max_speed <float>: Sets the maximum average speed in meters per second for a sequence of points to be considered a low-speed segment. Default: 3.0.
• --min_time <float>: Sets the minimum time duration in seconds for a low-speed segment to be considered valid. Default: 60.0.
• --clear_existing: If provided, all previously stored DetectedSegment records will be deleted before processing new ones.

Example Usage:

To process trajectories using a maximum speed of 2.5 m/s and a minimum segment duration of 90 seconds:

python manage.py process_trajectories --max_speed 2.5 --min_time 90

To process trajectories only for vehicle IDs 101 and 102:

python manage.py process_trajectories --vehicle_ids 101 102

2. report_segments

This command generates a statistical report based on the DetectedSegment data stored in the database. It provides insights into the characteristics and distribution of detected low-speed segments.

Statistics Provided:

• Overall averages: Total segments, average duration, average speed, average number of points per segment, and total distance covered by segments.
• Top N routes: Lists routes with the highest number of detected segments.
• Hourly distribution: Shows the number of segments that started in each hour of the day.

Key Command-Line Arguments:

• --top_n_routes <int>: Customizes the number of top routes to display in the report. Default: 5.

Example Usage:

To generate a report showing the top 3 routes by segment count:

python manage.py report_segments --top_n_routes 3

To generate a report with default settings:

python manage.py report_segments

Typical Workflow

1. Data Population: Ensure that vehicle trajectory data is loaded into the system's Trajectory model. (This step is typically handled by other processes, e.g., populate_database or run_service_tracks based on existing commands).
2. Process Trajectories: Run the process_trajectories command to analyze the raw trajectories. This will identify low-speed segments and populate the DetectedSegment table in the database.

   python manage.py process_trajectories --max_speed 2.0 --min_time 120

3. Generate Report: After processing, run the report_segments command to generate statistical insights from the newly populated DetectedSegment data.

   python manage.py report_segments --top_n_routes 10

This workflow allows users to transform raw GPS tracks into actionable insights about low-speed events, potentially indicating congestion or operational stops, and then quantify these events through statistical summaries.