Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction

Traffic accident prediction is essential for improving road safety and optimizing intelligent transportation systems. However, deep learning models often struggle with distribution shifts and class imbalance, leading to degraded performance in real-world applications. While distribution shift is a c...

Full description

Saved in:
Bibliographic Details
Main Authors: Duo Peng, Weiqi Yan
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Computers
Subjects:
Online Access:https://www.mdpi.com/2073-431X/14/5/186
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849711163239038976
author Duo Peng
Weiqi Yan
author_facet Duo Peng
Weiqi Yan
author_sort Duo Peng
collection DOAJ
description Traffic accident prediction is essential for improving road safety and optimizing intelligent transportation systems. However, deep learning models often struggle with distribution shifts and class imbalance, leading to degraded performance in real-world applications. While distribution shift is a common challenge in machine learning, Transformer-based models—despite their ability to capture long-term dependencies—often lack mechanisms for dynamic adaptation during inferencing. In this paper, we propose a TTT-Enhanced Transformer that incorporates Test-Time Training (TTT), enabling the model to refine its parameters during inferencing through a self-supervised auxiliary task. To further boost performance, an Adaptive Memory Layer (AML), a Feature Pyramid Network (FPN), Class-Balanced Attention (CBA), and Focal Loss are integrated to address multi-scale, long-term, and imbalance-related challenges. Our experimental results show that our model achieved an overall accuracy of 96.86% and a severe accident recall of 95.8%, outperforming the strongest Transformer baseline by 5.65% in accuracy and 9.6% in recall. The results of our confusion matrix and ROC analyses confirm our model’s superior classification balance and discriminatory power. These findings highlight the potential of our approach in enhancing real-time adaptability and robustness under shifting data distributions and class imbalances in intelligent transportation systems.
format Article
id doaj-art-dfadb9a1af084890bddd80779e927884
institution DOAJ
issn 2073-431X
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Computers
spelling doaj-art-dfadb9a1af084890bddd80779e9278842025-08-20T03:14:41ZengMDPI AGComputers2073-431X2025-05-0114518610.3390/computers14050186Test-Time Training with Adaptive Memory for Traffic Accident Severity PredictionDuo Peng0Weiqi Yan1Department of Computer and Information Science, Auckland University of Technology, Auckland 1010, New ZealandDepartment of Computer and Information Science, Auckland University of Technology, Auckland 1010, New ZealandTraffic accident prediction is essential for improving road safety and optimizing intelligent transportation systems. However, deep learning models often struggle with distribution shifts and class imbalance, leading to degraded performance in real-world applications. While distribution shift is a common challenge in machine learning, Transformer-based models—despite their ability to capture long-term dependencies—often lack mechanisms for dynamic adaptation during inferencing. In this paper, we propose a TTT-Enhanced Transformer that incorporates Test-Time Training (TTT), enabling the model to refine its parameters during inferencing through a self-supervised auxiliary task. To further boost performance, an Adaptive Memory Layer (AML), a Feature Pyramid Network (FPN), Class-Balanced Attention (CBA), and Focal Loss are integrated to address multi-scale, long-term, and imbalance-related challenges. Our experimental results show that our model achieved an overall accuracy of 96.86% and a severe accident recall of 95.8%, outperforming the strongest Transformer baseline by 5.65% in accuracy and 9.6% in recall. The results of our confusion matrix and ROC analyses confirm our model’s superior classification balance and discriminatory power. These findings highlight the potential of our approach in enhancing real-time adaptability and robustness under shifting data distributions and class imbalances in intelligent transportation systems.https://www.mdpi.com/2073-431X/14/5/186test-time trainingtraffic accident predictiontransformer networkself-supervised learningadaptive memoryclass-balanced learning
spellingShingle Duo Peng
Weiqi Yan
Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
Computers
test-time training
traffic accident prediction
transformer network
self-supervised learning
adaptive memory
class-balanced learning
title Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
title_full Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
title_fullStr Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
title_full_unstemmed Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
title_short Test-Time Training with Adaptive Memory for Traffic Accident Severity Prediction
title_sort test time training with adaptive memory for traffic accident severity prediction
topic test-time training
traffic accident prediction
transformer network
self-supervised learning
adaptive memory
class-balanced learning
url https://www.mdpi.com/2073-431X/14/5/186
work_keys_str_mv AT duopeng testtimetrainingwithadaptivememoryfortrafficaccidentseverityprediction
AT weiqiyan testtimetrainingwithadaptivememoryfortrafficaccidentseverityprediction