Vehicle Type-Specific Speed Prediction Model

The patent segments the mixed vehicle type tracking data into separate datasets based on vehicle types. The system identifies and separates tracking data for different vehicle types (cars, motorcycles, bicycles, public transit) from the mixed data stream, creating dedicated training datasets for each vehicle type. This segmentation enables the machine learning model to learn vehicle-type-specific speed patterns while maintaining adequate data coverage through the use of multiple separated datasets.

The patent changes the parameter of speed estimation from generic to vehicle-type-specific by incorporating vehicle type as a key parameter in the machine learning model. The system maintains relatively low complexity by using a supervised learning approach where the model learns to predict speeds based on vehicle type and road segment features. This parameter change significantly improves travel time estimation accuracy while keeping the system manageable through the use of standard machine learning techniques.

The patent implements self-service by automatically identifying vehicle types from tracking data and training separate machine learning models for each vehicle type without requiring manual intervention. The system autonomously processes mixed vehicle type data, separates it by vehicle type, trains appropriate models, and provides vehicle-type-specific route recommendations. This self-service approach maintains ease of operation for users while dramatically improving route optimization effectiveness through specialized speed predictions.