Driver driving behavior recognition and classification method and system based on driving feature group

Abstract

The invention provides a driver driving behavior recognition and classification method and system based on a driving feature group, and belongs to the technical field of data analysis. The method comprises the steps of collecting the driving track data of multiple drivers in a fixed time period from a database to serve as an original data set, performing the feature extraction on the driving tracks of the drivers in the fixed time period according to the road traffic safety rules to obtain the driving behavior features of the drivers, normalizing the extracted driving behavior characteristicsof the drivers to obtain the driving characteristic vectors, performing principal component analysis dimension reduction on the driving feature vector, according to the driving feature vectors after dimension reduction, combining a mean value clustering algorithm k-means to identify and classify the driving behaviors. According to the method, ten kinds of driving characteristics reflecting the driving behavior tendency of the drivers are comprehensively extracted, the drivers are divided into a dangerous type, a common type and a mild type through the kmeans clustering, and the method can be used for a road traffic safety supervision department, a vehicle-mounted voice system and the like to warn the drivers with the potential safety hazards.

Description

technical field [0001] The invention relates to the technical field of data analysis, in particular to a method and system for identifying and classifying drivers' driving behaviors based on driving feature groups. Background technique [0002] Data mining is an intelligent means of extracting effective decision-making information and knowledge from massive data. Traditional data mining tasks mainly include classification analysis, cluster analysis and association rule discovery. For clustering analysis (ClusteringAnalysis), clustering is to divide the database records without category marks into several disjoint subsets (clusters) according to a given similarity metric, so that the internal records of each cluster are similar The degree of similarity between different clusters is very low. Properly extracting the features of the category to be classified is the most critical link in this unsupervised clustering without category labels. [0003] At present, the evaluation...

AI Technical Summary

Problems solved by technology

[0003] At present, the evaluation of drivers is mainly for the identification of detailed driving behaviors, and experts score the identified driving behavior units. The problem is that if the types of identified driving behaviors are too few, the evaluation of driving behaviors is not comprehensive and accurate enough. , the coverage ability is poor. If there are too many types of driving behaviors identified and used for scoring, it will be more difficult for experts to score, and the influence of each behavior on the scoring results will be small, resulting in a decrease in accuracy

Images