Hybrid Departure Time Prediction Using Clustering and Association Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting departure times in automotive applications are inadequate, particularly in scenarios where users spend considerable time near their vehicles, as they require more time for proactive personalized services like preconditioning and are challenged by limited availability of Bluetooth or WiFi signals.
Innovation Solution
A hybrid method and system that combines trip/data processing, predictive modeling, and hybrid predictive departure time determination, using clustering of stay points and contextual association rule mining to infer user behavior and patterns from historical and context data, allowing for improved prediction of departure times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Bluetooth or WiFi signal strength detection is used to predict departure time, then the system can detect user proximity, but the prediction accuracy deteriorates when users spend considerable time near the vehicle
Solution Approach 1:
The patent segments the prediction approach into multiple independent components: signal-based detection, behavior-based prediction, and pattern recognition. Each component handles specific scenarios independently, allowing the system to switch between them based on context. This segmentation resolves the contradiction by providing specialized solutions for different user proximity scenarios without relying on a single flawed approach.
Solution Approach 2:
The system changes the parameters used for prediction based on the scenario. When signal-based detection is insufficient (users spending considerable time near vehicle), the system transitions to using behavioral parameters and historical patterns. This parameter change allows accurate prediction in scenarios where traditional signal strength detection fails.
2Loss of time
If behavior-based prediction is used to predict departure time, then the system can provide proactive services, but the overhead time increases due to limited signal availability
Solution Approach 1:
The system performs preliminary action by continuously learning and storing user behavioral patterns and departure time preferences in advance. This pre-learned information is readily available when prediction is needed, eliminating the need for real-time signal detection and reducing overhead time while maintaining reliable proactive service delivery.
Solution Approach 2:
The system creates a behavioral model that copies and stores historical user patterns. Instead of relying on real-time signal detection, the system uses this copied behavioral data to predict departure times, significantly reducing overhead time while maintaining prediction reliability for proactive services.
3Measurement precision
If traditional prediction methods are used, then the system structure remains simple, but the prediction accuracy deteriorates for personalized services
Solution Approach 1:
The complex prediction system is segmented into modular components: data collection module, pattern recognition module, prediction module, and service execution module. Each module performs a specific function independently, making the overall complex system manageable and maintainable while achieving high prediction accuracy for personalized services.
Solution Approach 2:
The prediction system is designed with multi-functionality to handle various prediction scenarios (departure time, arrival time, stay duration) using a unified behavioral analysis framework. This universal approach improves prediction accuracy across different services while avoiding the need for multiple separate complex systems.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method and system for determining a predicted departure time of a user of a vehicle is disclosed. The method comprises determining, based on a user data set indicative of a plurality of trips made by the user: first data indicative of trip segment information, second data indicative of trip destinations, and third data indicative of semantic trip properties, determining, based on the first, second, and/or third data: fourth data indicative of a destination arrival behavior, fifth data indicative of a stay duration at a destination, and sixth data indicative of a departure behavior, determining, based on the fourth, fifth, and/or sixth data, the predicted departure time of the user. A system comprising a control unit configured for performing the method and a vehicle comprising the system are also disclosed.