Analysis on Train Stopping Accuracy based on Regression Algorithms

Stopping accuracy is one of the most important indexes of efficiency of automatic train operation (ATO) systems. Traditional stopping control algorithms in ATO systems have some drawbacks, as many factors have not been taken into account. In the large amount of fieldcollected data about stopping accuracy there are many factors (e.g. system delays, stopping time, net pressure) which affecting stopping accuracy. In this paper, three popular data mining methods are proposed to analyze the train stopping accuracy. Firstly, we find fifteen factors which have impact on the stopping accuracy. Then, ridge regression, lasso regression and elastic net regression are employed to mine models to reflecting the relationship between the fifteen factors and the stopping accuracy. Then, the three models are compared by using Akaike information criterion (AIC), a model selection criterion which considering the trade-off between accuracy and complexity. The computational results show that elastic net regression model has a best performance on AIC value. Finally, we obtain the parameters which can make the train stop more accurately which can provide a reference to improve stopping accuracy for ATO systems.