Road traffic state distinguishing method based on semi-supervised learning

Abstract

The invention provides a road traffic state distinguishing method based on semi-supervised learning. The method includes the steps of a) collecting traffic flow data of a road and upstream and downstream intersections, b) marking traffic states, c) generating a marker training set F1, d) generating marker training sets F2 and F3, e) training three kinds of base classifiers by means of the F1, the F2 and the F3, f) selecting m unmarked samples, g) carrying out cooperative training on the base classifiers, and h) repeatedly executing the step g) until cooperative training is finished. According to the road traffic state distinguishing method, the classifiers are trained by means of marked traffic flow data firstly, then cooperative training is conducted by means of unmarked traffic flow data, the unmarked cost-saving data are made full use of, more accurate base classifiers can be formed, traffic state distinguishing accuracy is improved, and traffic management, decision-making, planning, operation correctness and quality of intelligent transportation information service are promoted.

Description

technical field [0001] The present invention relates to a road traffic state discrimination method based on semi-supervised learning, more specifically, it relates to a method for firstly using marked traffic flow data to train a base classifier and then using unmarked traffic flow data for collaborative Trained semi-supervised learning based road traffic state discriminant method. Background technique [0002] With the rapid development of cities, traffic congestion and traffic pollution are becoming more and more serious. Traffic exhaust emission is one of the main pollution sources of smog. These are the urgent problems to be solved in major cities, and intelligent transportation has become the key to improving urban transportation. In recent years, the intelligent transportation systems of major and medium-sized cities have been continuously improved, and the degree of intelligence has been continuously improved. The scope, breadth and depth of traffic data collection h...

AI Technical Summary

Problems solved by technology

Even so, there are still many traffic state discrimination methods that are difficult to apply in practice. The main reason is that these supervised traffic state discrimination methods require a large number of training samples, and it is difficult to collect a large number of traffic state samples in practice, thus limiting many Application of Data Mining Method in Real Traffic Status Discrimination

[0004] If only a small number of labeled samples are used, it is often difficult for the learning system trained by them to have strong generalization ability; on the other hand, if only a small number of "expensive" labeled samples are used without using a large Unlabeled samples are a great waste of data resources

If the large-scale traffic flow data is fully utilized, the correct rate of traffic state discrimination will be improved, but the calculation amount of conventional traffic state discrimination methods will increase exponentially with the increase of sample size.

Images