Prediction of Train Arrival Delay Using Hybrid ELM-PSO Approach

Abstract

In this study, a hybrid method combining extreme learning machine (ELM) and particle swarm optimization (PSO) is proposed to forecast train arrival delays that can be used for later delay management and timetable optimization. First, nine characteristics (e.g., buffer time, the train number, and station code) associated with train arrival delays are chosen and analyzed using extra trees classifier. Next, an ELM with one hidden layer is developed to predict train arrival delays by considering these characteristics mentioned before as input features. Furthermore, the PSO algorithm is chosen to optimize the hyperparameter of the ELM compared to Bayesian optimization and genetic algorithm solving the arduousness problem of manual regulating. Finally, a case is studied to confirm the advantage of the proposed model. Contrasted to four baseline models (k-nearest neighbor, categorical boosting, Lasso, and gradient boosting decision tree) across different metrics, the proposed model is demonstrated to be proficient and achieve the highest prediction accuracy. In addition, through a detailed analysis of the prediction error, it is found that our model possesses good robustness and correctness.

A review of train delay prediction approaches

“…Furthermore, probability distributions for delays can always be reduced to a single-value prediction (e.g., the mean or quantiles). If the influence of the most important factors affecting train delays is precisely known, deterministic predictions should be the preferred choice of model as they generally produce higher prediction accuracy (Bao et al, 2021).…”

“…We can observe that NN have mostly been applied to predict the delay of passenger trains with a recent focus on real-time utilisations. They currently provide the most accurate prediction results (Bao et al, 2021), although they do neither need nor provide insights into the dynamics of the railway system.…”

Analyzing Factors Contributing to Real-time Train Arrival Delays using Seemingly Unrelated Regression Models

“…Train event variables refer to variables that keep track of both scheduled train activities and actual train movements in real practice. Many studies found that train event variables are the primary factors affecting train delays (Huang et al, 2020a;Lee et al, 2016;Taleongpong et al, 2022;Bao et al, 2021;Shi et al, 2021). Several studies revealed that current arrival delays are the most significant factors when predicting one-station ahead train delays (Bao et al, 2021;Shi et al, 2021), but Li et al (2021) indicated that this is only true if the current delays are longer than three minutes.…”

“…Many studies found that train event variables are the primary factors affecting train delays (Huang et al, 2020a;Lee et al, 2016;Taleongpong et al, 2022;Bao et al, 2021;Shi et al, 2021). Several studies revealed that current arrival delays are the most significant factors when predicting one-station ahead train delays (Bao et al, 2021;Shi et al, 2021), but Li et al (2021) indicated that this is only true if the current delays are longer than three minutes. Many existing studies have demonstrated good performance of the train delay prediction models when the upstream stations' delays and upstream trains' delays are considered (Li et al, 2022;Huang et al, 2020a,b;Oneto et al, 2018Oneto et al, , 2017.…”

Prediction of Real-Time Train Arrival Times Along the Swedish Southern Mainline

“…Oneto et al [16] employed extreme learning machines (ELM) together with train operation data and weather data to build a dynamic train delay prediction model. Li et al [3] and Bao et al [17] utilised particle swarm optimization algorithm to optimize hyperparameter of ELM when predicting train delay for real-time train dispatching. To comprehensively account for the temporal and spatial dependence between multiple trains and routes, Zhang et al [18] proposed a train spatiotemporal graph convolutional network to predict the collective cumulative effect of train delays.…”