Denoising traffic collision data using ensemble empirical mode decomposition (EEMD) and its application for...

Area of Science:

• Transportation Science
• Data Science
• Signal Processing

Background:

• Accurate traffic collision data analysis is crucial for resource allocation and safety investigations.
• Existing denoising methods often require a known list of high collision concentration locations (HCCL), which is frequently unavailable.
• Noise in traffic data can lead to false positives, wasting resources on unnecessary safety assessments.

Purpose of the Study:

• To develop an innovative approach for denoising traffic collision data without relying on pre-existing HCCL lists.
• To introduce and apply the Ensemble Empirical Mode Decomposition (EEMD) method for traffic data analysis.
• To construct Continuous Risk Profiles (CRPs) from denoised data for better safety assessment.

Main Methods:

• Traffic collision data was transformed for decomposition using the Ensemble Empirical Mode Decomposition (EEMD) method.
• The EEMD method decomposed the data into Intrinsic Mode Functions (IMFs) and a residue.
• Intrinsic Mode Functions (IMFs) were analyzed to filter noise and construct Continuous Risk Profiles (CRPs).

Main Results:

• The study successfully generated denoised Continuous Risk Profiles (CRPs) using the EEMD method.
• The performance of the EEMD-based denoising was compared against traditional weighted moving window techniques.
• The developed method provides a viable alternative for denoising traffic data when HCCL lists are absent.

Conclusions:

• The Ensemble Empirical Mode Decomposition (EEMD) method offers an effective solution for denoising traffic collision data.
• This approach enhances the reliability of safety assessments and optimizes resource allocation for government agencies.
• Further research is recommended to refine the method and explore its application in diverse traffic safety contexts.