Traffic Forecasting Using Matrix Compression and Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current traffic prediction techniques face challenges in efficiently processing and analyzing large, sparse Origin-Destination matrices, which are computationally costly and inefficient, especially when dealing with granular traffic data, and struggle to accurately model the impact of special events on traffic conditions.
Innovation Solution
The method involves compressing Origin-Destination matrices using locality-aware techniques such as space-filling curves to maintain only non-empty data, and then using a trained Deep Neural Network to predict traffic conditions by learning from past observations, effectively reducing computational effort and preserving data locality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional traffic prediction techniques process complete Origin-Destination matrices, then comprehensive traffic data is analyzed, but computational cost and processing time increase significantly
Solution Approach 1:
The patent extracts and retains only non-empty data elements from the Origin-Destination matrix, removing redundant zero-value entries. This selective extraction reduces the data volume requiring processing while preserving all meaningful traffic information, thereby decreasing processing time without sacrificing prediction accuracy.
Solution Approach 2:
The patent applies locality-aware compression techniques that preserve the spatial relationships and geographical locality of traffic data. By maintaining the local structure of non-empty data elements and their relative positions, the system ensures that computational efficiency is improved while the accuracy of local traffic pattern analysis remains intact.
2Measurement precision
If granular traffic data is processed in detail, then prediction accuracy improves, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential non-empty data elements from granular traffic matrices, eliminating redundant computations on zero-value entries. This selective processing maintains the detailed information necessary for accurate predictions while reducing the overall computational complexity by focusing resources on meaningful data only.
Solution Approach 2:
The patent transforms the representation of traffic data by changing from a complete matrix format to a compressed format that stores only non-empty elements with their positional information. This parameter change in data representation reduces computational complexity while preserving the granular detail needed for accurate prediction through the use of space-filling curves and locality-aware encoding.
3Loss of information
If complete Origin-Destination matrices are stored and processed, then all traffic information is available, but data redundancy and storage requirements increase
Solution Approach 1:
The patent extracts and stores only non-empty data elements from the Origin-Destination matrix, removing all redundant zero-value entries. This extraction process eliminates data redundancy while preserving complete traffic information, as the system retains all meaningful traffic flow data along with their spatial relationships through locality-aware compression.
Solution Approach 2:
The patent preserves the local structure and spatial relationships of traffic data through locality-aware compression techniques. By maintaining the relative positions and geographical context of non-empty data elements, the system ensures that no meaningful traffic information is lost while significantly reducing storage requirements by eliminating redundant zero-value entries.
Data Source
AI summary
Traffic condition forecasting techniques are provided that use matrix compression and deep neural networks. An illustrative method comprises obtaining a compressed origination-destination matrix indicating a cost to travel between pairs of a plurality of nodes, wherein the compressed origination-destination matrix is compressed using a locality-aware compression technique that maintains only non-empty data; obtaining a trained deep neural network trained using the compressed origination-destination matrix and past observations of traffic conditions at various times; and applying traffic conditions between two nodes in the compressed origination-destination matrix at a time, t, to the trained deep neural network to obtain predicted traffic conditions between the two nodes at a time, t+Δ. A tensor can be generated indicating an evolution of traffic conditions over a time span using a stacked Origination-Destination matrix comprising a plurality of past observations representing the time span.