Data pertaining to Chapter 6 "Human Driving Patterns - A Knowledge-Enhanced Deep Learning Approach for Behaviour Modelling"

This dataset accompanies the paper <em>“Human Driving Patterns – A Knowledge-Enhanced Deep Learning Approach for Behaviour Modelling”</em> (Chapter 6 of the PhD dissertation). The research focuses on data-driven modelling of longitudinal driving behaviour using a novel knowledge-enhanced deep learning framework. It aims to integrate expert knowledge with deep learning to improve the interpretability and accuracy of driver behaviour models. A Knowledge-Enhanced Attention LSTM (KE-ALSTM) model to predict transitions and durations of Action patterns. Graph-based and distribution-based knowledge are integrated to improve DL model performance. Evaluation of real-world data demonstrates that KE-ALSTM outperforms baseline models, demonstrating the value of incorporating domain knowledgeto enhance deep-learning models in driving behaviour analysis. The dataset was created and processed through a combination of data preprocessing, feature extraction, and model training in MATLAB and Python. It is provided as a zipped folder containing files in <code>.xlsx</code>, <code>.csv</code>, <code>.mat</code>, <code>.m</code>, <code>.txt</code>, and <code>.pdf</code> formats. A <code>ch6_Readme.txt</code> file is included to guide users on how to access and use the data for reproduction and further research.本数据集配套于论文《人类驾驶模式——一种用于行为建模的知识增强深度学习方法》（该论文为博士学位论文第6章）。本研究依托全新的知识增强深度学习框架，聚焦于纵向驾驶行为的数据驱动建模，旨在将专家知识与深度学习相融合，以提升驾驶员行为模型的可解释性与预测精度。本研究采用知识增强注意力长短期记忆网络（Knowledge-Enhanced Attention LSTM，KE-ALSTM）模型，实现对行为模式的切换过程与持续时长的预测；同时整合基于图结构与基于分布的两类知识，以提升深度学习（Deep Learning，DL）模型的性能。基于真实世界采集数据的评估结果表明，KE-ALSTM模型的性能优于基准模型，印证了在驾驶行为分析领域引入领域知识以优化深度学习模型的应用价值。本数据集通过结合MATLAB与Python工具完成数据预处理、特征提取及模型训练等环节制作并处理完成，以压缩包形式提供，内含.xlsx、.csv、.mat、.m、.txt及.pdf格式的各类文件。压缩包中还包含ch6_Readme.txt文件，用于指导用户如何获取并使用本数据集以完成实验复现与后续研究工作。