Improve Aggressive Driver Recognition Using Collision Surrogate Measurement and Imbalanced Class Boosting

Abstract

Real-time recognition of risky driving behavior and aggressive drivers is a promising research domain, thanks to powerful machine learning algorithms and the big data provided by in-vehicle and roadside sensors. However, since the occurrence of aggressive drivers in real traffic is infrequent, most machine learning algorithms treat each sample equally and prone to better predict normal drivers rather than aggressive drivers, which is our real interest. This paper aims to test the advantage of imbalanced class boosting algorithms in aggressive driver recognition using vehicle trajectory data. First, a surrogate measurement of collision risk, called Average Crash Risk (ACR), is proposed to calculate a vehicle’s crash risk. Second, the driver’s driving aggressiveness is determined by his/her ACR with three anomaly detection methods. Third, we train classification models to identify aggressive drivers using partial trajectory data. Three imbalanced class boosting algorithms, SMOTEBoost, RUSBoost, and CUSBoost, are compared with cost-sensitive AdaBoost and cost-sensitive XGBoost. Additionally, we try two resampling techniques with AdaBoost and XGBoost. Among all algorithms tested, CUSBoost achieves the highest or the second-highest Area Under Precision-Recall Curve (AUPRC) in most datasets. We find the discrete Fourier coefficients of gap as the key feature to identify aggressive drivers.

Risky Driver Recognition with Class Imbalance Data and Automated Machine Learning Framework

“…Future efforts should focus on the following aspects: (1) a more comprehensive collision risk evaluation on the target vehicle is needed to establish more reliable ground truth. (2) The framework proposed in this paper can be extended to other machine learning algorithms, such as deep neural networks, bagging and stacking of XGBoost classifiers.…”

“…When DSS < 0, the following vehicle has a collision risk. Wang [ 2 ] proposed measurement of Collision Risk ( CR ) as the absolute value of DSS divided by the following vehicle’s speed. …”

“…Most research on risky driving and risky driver recognition algorithms focuses on risky driving state recognition, including aggressive driving, distracted driving, fatigue driving, etc. For example, Wang et al [ 2 ] used discrete Fourier coefficients of vehicle trajectory data, such as distance between vehicles and speed, as input and used imbalanced class boosting algorithms to identify aggressive car-following drivers. Sun et al [ 3 ] combined background features such as driving time and sleep time to establish a dual-layer fusion fatigue driving recognition model based on the driver’s facial features and operating characteristics.…”

“…Some studies also use a driving simulator to observe the driving behavior of experimenters in a pre-designed driving environment [ 13 , 14 ]. In addition, cameras deployed on the roadside or unmanned aerial vehicles (UAV) can record traffic video, from which advanced computer vision algorithms can extract vehicle trajectory data at a lower cost compared to naturalistic driving experiment and driving simulator [ 2 , 10 ]. A smartphone is equipped with multiple sensors, including an accelerometer, gyroscope, magnetometer, microphone, cameras, thermometer, and Global Positioning System (GPS).…”

Research on Imbalanced Microscopic Image Classification of Harmful Algae

“…This not only can eliminate the highlight background region formed by dense tissues, but also can retain regions which have similar shape and features with the cingulum. In this section, the dominant gray levels set of images were acquired by a grayscale inclusion sphere [20]. Meaningless local minima regions can be decreased by decreasing number of gray levels in images effectively.…”