Driven by the era of big data and the development of artificial intelligence, potential traffic patterns can be obtained by analysing the numerous data. Metro has become an essential transport infrastructure and the passenger volume provides the basic support for the optimisation of the metro system. Thus, accurate forecasting of the volume is extremely required. In this study, a model for improving the accuracy and stability of metro passenger volume prediction named VMD‐TPE‐LightGBM (light gradient boosting machine) is proposed. The original dataset is firstly regrouped both in the station and chronological order while the time interval is reset as 10‐minute. Time features for extracting the hidden patterns are extracted by analysing the variation tendency of the passenger volume. For enhancing the precision, the variational mode decomposition algorithm is applied to decompose the original data series. Then each of the modes is regarded as the input of the LightGBM model, which are optimised by a tuning method named the tree of Parzen estimators and K‐fold cross‐validation. According to this process, the final forecasting results are acquired by reconstructing the predicted modes. The experimental results demonstrate that the proposed model performs superior to all the comparisons and has an impressive effect on short‐term metro passenger volume forecasting.
1 January 2016 in Mathematical Problems in Engineering
In order to coordinate the complex relationship between supplies distribution and path selection, some influential factors must be taken into account such as the insufficient remaining capacity of the road and uncertainty of travel time during supplies distribution and transportation. After the structure of emergency logistics network is analyzed, the travel time Bayes risk function of path and the total loss Bayes risk function of the disaster area are proposed. With the emergency supplies total transportation unit loss as the goal, an emergency logistics network optimization model under crowded conditions is established by the Bayes decision theory and solved by the improved ant colony algorithm. Then, a case of the model is validated to prove that the emergency logistics network optimization model is effective in congested conditions.
5 March 2019 in Journal of Advanced Transportation
We study the problem of path choice for emergency logistics in this paper. Based on the uncertainty environment during the path choice from emergency logistics network and the bounded rationality of decision makers, cumulative prospect theory is introduced to study the problem of emergency logistics path choice with comprehensive consideration of path properties and risk attitude of decision makers. In addition, the decision behavior of decision maker with the attitude of risk seeking and risk aversion under limited rationality is comprehensively analyzed respectively. Based on the choice behavior, a strategy to demarcate the value of reference point value is also proposed, and an optimization model is used to obtain the combined weight based on the moment estimation. Finally, both the theory and model are verified by calculation and compared analysis in a case study. In addition, perturbation analyses of related parameter are carried out to further reveal the influence mechanism between the prospect value of each path and related parameters. The result shows that the decision-making model can make emergency logistics path choice with higher efficiency and reliability under different complex interference conditions.
4 March 2019 in Journal of Advanced Transportation
This paper focuses on the discrete robustness optimization of emergency transportation network with the consideration of timeliness and decision behavior of decision-maker under the limited rationality. Based on a situation that the nearer to disaster area, the higher probability of time delay, prospect theory is specially introduced to reflect the subjective decision behavior of decision-maker. Then, a discrete robustness optimization model is proposed with the purpose of the better timeliness and robustness. The model is based on the emergency transportation network with multistorage centers and multidisaster points. In order to obtain the optimal solution, an improved genetic algorithm is designed by introducing a bidirectional search strategy based on a newfangled path cluster to obtain specific paths that connect each storage centers and each disaster points. Finally, a case study is exhibited to demonstrate the reasonability of the model, theory, and algorithm. The result shows that the path cluster with the better timeliness and robustness can be well obtained by using the prospect theory and improved genetic algorithm. The analysis especially reveals that the robustness is correspondent to the risk aversion in prospect theory.
24 November 2020 in Journal of Advanced Transportation
This paper focuses on discrete dynamic optimization on train rescheduling on single-track railway with the consideration of train punctuality and station satisfaction degree. A discrete dynamic system is firstly described to mimic train rescheduling, and a state transition function is specially designed according to the train departure event. The purpose of this function is to improve simulation efficiency by directly confirming the next discrete time. After the construction and analysis of optimization models to discrete dynamic system, a two-stage heuristic search strategy is developed, by using clustering hierarchy theory and stochastic search strategy, to obtain train departure time and arrival time before each state transition of the system. Finally, a numerical experiment is conducted to verify the proposed system, models, and the heuristic search strategy. The result shows that the discrete dynamic system, together with the state transition function and heuristic search strategy, shows better performance in simulation efficiency and solution quality.
26 June 2023 in Proceedings of the AAAI Conference on Artificial Intelligence
Neural machine translation (NMT) has achieved remarkable success in producing high-quality translations. However, current NMT systems suffer from a lack of reliability, as their outputs that are often affected by lexical or syntactic changes in inputs, resulting in large variations in quality. This limitation hinders the practicality and trustworthiness of NMT. A contributing factor to this problem is that NMT models trained with the one-to-one paradigm struggle to handle the source diversity phenomenon, where inputs with the same meaning can be expressed differently. In this work, we treat this problem as a bilevel optimization problem and present a consistency-aware meta-learning (CAML) framework derived from the model-agnostic meta-learning (MAML) algorithm to address it. Specifically, the NMT model with CAML (named CoNMT) first learns a consistent meta representation of semantically equivalent sentences in the outer loop. Subsequently, a mapping from the meta representation to the output sentence is learned in the inner loop, allowing the NMT model to translate semantically equivalent sentences to the same target sentence. We conduct experiments on the NIST Chinese to English task, three WMT translation tasks, and the TED M2O task. The results demonstrate that CoNMT effectively improves overall translation quality and reliably handles diverse inputs.