Abstract:Early risk diagnosis and driving anomaly detection from vehicle stream are of great benefits in a range of advanced solutions towards Smart Road and crash prevention, although there are intrinsic challenges, especially lack of ground truth, definition of multiple risk exposures. This study proposes a domain-specific automatic clustering (termed Autocluster) to self-learn the optimal models for unsupervised risk assessment, which integrates key steps of risk clustering into an auto-optimisable pipeline, including feature and algorithm selection, hyperparameter auto-tuning. Firstly, based on surrogate conflict measures, indicator-guided feature extraction is conducted to construct temporal-spatial and kinematical risk features. Then we develop an elimination-based model reliance importance (EMRI) method to unsupervised-select the useful features. Secondly, we propose balanced Silhouette Index (bSI) to evaluate the internal quality of imbalanced clustering. A loss function is designed that considers the clustering performance in terms of internal quality, inter-cluster variation, and model stability. Thirdly, based on Bayesian optimisation, the algorithm selection and hyperparameter auto-tuning are self-learned to generate the best clustering partitions. Various algorithms are comprehensively investigated. Herein, NGSIM vehicle trajectory data is used for test-bedding. Findings show that Autocluster is reliable and promising to diagnose multiple distinct risk exposures inherent to generalised driving behaviour. Besides, we also delve into risk clustering, such as, algorithms heterogeneity, Silhouette analysis, hierarchical clustering flows, etc. Meanwhile, the Autocluster is also a method for unsupervised multi-risk data labelling and indicator threshold calibration. Furthermore, Autocluster is useful to tackle the challenges in imbalanced clustering without ground truth or priori knowledge

Clustering Driver Behavior Using Dynamic Time Warping and Hidden Markov Model

Dynamic Driving Intention Recognition of Vehicles with Different Driving Styles of Surrounding Vehicles

Discovery of Driving Behavior Patterns Based on Multivariate Time Series

Clustering of Several Typical Behavioral Characteristics of Commercial Vehicle Drivers Based on GPS Data Mining: Case Study of Highways in China

Understanding Drivers’ Travel Behaviors Through Vehicle Onboard Diagnostic Data Using Multi-Dimensional Discrete Hidden Markov Model

Unsupervised Driving Style Analysis Based on Driving Maneuver Intensity.

A dynamic driving-style analysis method based on drivers' interaction with surrounding vehicles

Clustering Analysis of Typical Ramp Scenarios Based on Naturalistic Driving Data

Driving Style Recognition of Taxi Drivers Based on Naturalistic Driving Data

Driving Behavior Clustering for Hazardous Material Transportation Based on Genetic Fuzzy C-Means Algorithm

Characterizing driver behavior using naturalistic driving data

High-risk event prone driver identification considering driving behavior temporal covariate shift

Feature selection for driving style and skill clustering using naturalistic driving data and driving behavior questionnaire

Driving Behavior Evaluation Based on DBSCAN and Kmeans++ Clustering

A Rapid Pattern-Recognition Method for Driving Styles Using Clustering-Based Support Vector Machines

Clustering Vehicle Temporal and Spatial Travel Behavior Using License Plate Recognition Data

Automatic Clustering for Unsupervised Risk Diagnosis of Vehicle Driving for Smart Road

Factor analysis and fuzzy clustering of driver behavior pattern

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