In this paper, we develop a heuristic model based on Gaussian processes to determine synthetic sets of trips in urban networks, considering only supply-related information. This is an alternative to the benchmark method used in the literature, which consists of repeating several trials of Monte Carlo simulations and therefore requiring a complex calibration task and large computational resources. The developed heuristic model explicitly leverages the probabilistic nature of Gaussian processes and exploits their properties to iteratively select origin–destination (od) pairs of nodes in the city network. The model then determines the shortest trip in distance for the selected od pairs and appends it to the synthetic set. We discuss the implementation and performance of both the benchmark method and the developed heuristic model on two city networks. We show that the presented model is more robust and computationally efficient than the benchmark method. This is evidenced by its ability to determine synthetic sets with much smaller sizes, naturally reducing the computational burden, when compared to the benchmark method. We also discuss how the choice of the kernel function and calibration of the hyperparameters influence the performance of the presented heuristic model.

中文翻译：

---

用于开发合成行程集的高斯采样启发式估计模型

在本文中，我们开发了一种基于高斯过程的启发式模型，以确定城市网络中的综合出行集，仅考虑与供应相关的信息。这是文献中使用的基准方法的替代方法，该方法包括重复多次蒙特卡罗模拟试验，因此需要复杂的校准任务和大量计算资源。开发的启发式模型显式地利用了高斯过程的概率性质并利用它们的特性来迭代地选择城市网络中的起点-终点 (od) 节点对。然后，模型确定所选 od 对的最短距离行程并将其附加到合成集。我们讨论了基准方法和开发的启发式模型在两个城市网络上的实现和性能。我们表明，所提出的模型比基准方法更健壮且计算效率更高。与基准方法相比，它能够确定尺寸小得多的合成集，自然减少了计算负担，这证明了这一点。我们还讨论了核函数的选择和超参数的校准如何影响所呈现的启发式模型的性能。