Research on Driver Driving Style and Driving Condition Recognition Model Based on SVM and XGBoost Nanjing Univ of Science and Technology

Format:

Book

Conference/Event

Author/Creator:

Shi, Shaojian, author.

Contributor:

Ding, Yang

Qian, Yahui

Song, Wei

Wang, Tao

Zou, Xiaojun

Conference Name:

WCX SAE World Congress Experience (2022-04-05 : Detroit & Online, Michigan, United States)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2022

Summary:

At present, the remote monitoring cloud platform of many automobile companies only displays the collected data information, and it does not fully mine the deep-level information of the data. This paper uses data mining and machine learning methods to build a driver's driving style and driving condition prediction and recognition model based on the historical driving information generated by the vehicle, so as to improve the supervision and safety of the driver and the vehicle by automobile companies and other automobile-related industries. First, 36 standard driving cycles are utilized to construct an initial operating condition block data set. Second, we obtain the feature variables of driving style and driving conditions through feature engineering, and two recognition model data sets use the principal component analysis (PCA) and clustering algorithm for data dimensionality reduction and cluster analysis. Then, two types of supervised learning, support vector machine (SVM) and extreme gradient boosting model (XGBoost), are selected for training, 50% of the data in the data set is randomly selected to predict the two recognition models ten times. Results show that the average prediction accuracy and time of the driving style recognition model are 99.4% and 0.00239 s, and the average prediction accuracy and time of the driving condition recognition model are 99.2% and 0.03198 s, which have good predictive performance. The vehicle data in the company's online cloud platform is further applied for predictive verification, and the results demonstrate that the established model has high feasibility

Notes:

Vendor supplied data

Publisher Number:

2022-01-0227

Access Restriction:

Restricted for use by site license