A Multi-Clustering Algorithm to Solve Driving Cycle Prediction Problems Based on Unbalanced Data Sets: A Chinese Case...

Area of Science:

• Transportation Engineering
• Data Science
• Machine Learning

Background:

• Vehicle evaluation parameters are crucial for consumers and governments.
• Existing driving cycle prediction algorithms struggle with China's complex traffic and unbalanced datasets.
• Unbalanced data significantly impacts the performance of traditional clustering algorithms for driving cycle mapping.

Purpose of the Study:

• To propose a novel algorithm framework for driving cycle map creation.
• To address the challenges of unbalanced datasets in vehicle driving cycle prediction.
• To enhance the performance of clustering algorithms in complex traffic environments.

Main Methods:

• Developed a multi-clustering algorithm based on ensemble learning.
• Designed a flexible, modular model framework for easy extension.
• Tested the algorithm using real-world traffic data from Fujian Province, China.

Main Results:

• The multi-clustering algorithm demonstrated excellent performance on unbalanced datasets.
• The proposed method effectively improves driving cycle map accuracy.
• The framework is adaptable to other complex structured areas.

Conclusions:

• The multi-clustering algorithm offers a robust solution for driving cycle prediction with unbalanced data.
• This approach is particularly beneficial for complex traffic environments like those in China.
• The flexible design allows for broader applications in transportation data analysis.