Traffic Prediction Device Using Requirement and Time-Series Feature Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing traffic prediction methods, such as those using RNN and machine learning models, struggle to accurately predict traffic flow rates due to changes in line requirements and traffic variations over time, leading to large errors and reduced accuracy.
Innovation Solution
A traffic prediction device that acquires line requirements and traffic data, generates feature amounts representing traffic variations due to requirement changes and traffic statistics, and uses these features to predict future traffic flow rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If RNN is used for traffic prediction, then prediction can be performed based on past traffic data, but the prediction value cannot follow traffic variation when line requirements change, resulting in large prediction errors
Solution Approach 1:
The patent merges two prediction approaches: RNN-based time-series prediction and machine learning prediction from requirements. The prediction result is obtained by combining both prediction values, allowing the system to leverage both past traffic patterns and requirement changes, thereby resolving the contradiction between maintaining prediction accuracy and adapting to requirement changes.
Solution Approach 2:
The system incorporates feedback from both traffic data and requirement information to continuously improve prediction accuracy. By using the actual traffic flow rate as feedback to evaluate and adjust the prediction model, the system adapts to changing conditions while maintaining accuracy.
2Adaptability or versatility
If machine learning model predicts traffic from line requirements, then prediction can be performed for each requirement set, but it is impossible to follow traffic variations over time and accuracy is lowered when correlation between requirements and traffic is low
Solution Approach 1:
The patent combines machine learning prediction from requirements with RNN-based time-series prediction. This merging allows the system to capture both the impact of requirement changes and temporal traffic patterns, resolving the contradiction between adapting to requirement variations and maintaining prediction accuracy when correlation is low.
Solution Approach 2:
The prediction system uses a composite approach by integrating multiple prediction methods (RNN and machine learning) into a unified model. This composite prediction mechanism leverages the strengths of each individual method while compensating for their weaknesses, enabling accurate prediction even when the correlation between requirements and traffic is low.
3Measurement precision
If only traffic data is used for prediction, then time-series patterns can be captured, but changes in line requirements are not reflected in the prediction
Solution Approach 1:
The patent merges traffic data-based prediction with requirement-based prediction. By combining these two information sources, the system captures both temporal patterns from traffic data and structural changes from requirement modifications, thereby preventing information loss while maintaining prediction accuracy.
Solution Approach 2:
The prediction system segments the input information into two distinct components: traffic data for capturing temporal patterns and requirement information for detecting structural changes. This segmentation allows each component to be processed appropriately before being integrated into the final prediction, ensuring that no critical information is lost.
Data Source
AI summary
A traffic prediction device (1) according to the present invention includes a data acquisition unit (11) that acquires a requirement of the line and traffic data in the line, a prediction data generation unit (12) that generates a first feature amount corresponding to a traffic variation due to a change in a requirement of the line based on the acquired requirement, and generates a second feature amount based on traffic statistics representing a feature of the traffic variation for each time based on the acquired traffic data; and a prediction function unit (13) that predicts a future traffic flow rate from the first feature amount and the second feature amount.