Interpretable traffic flow prediction stacking method

The invention discloses an interpretable traffic flow prediction stacking method. The method comprises the following steps: S1, collecting spatial-temporal characteristics related to traffic flow prediction and preprocessing data; s2, building a stacking model, debugging and checking; and S3, evaluating the influence of each spatio-temporal feature on a prediction result through a consistency clustering thought. According to the invention, the accuracy of the prediction result is improved, and the interpretability of the prediction result is ensured; a new framework aiming at enhancing prediction precision and model interpretability is provided, so that traffic professionals and stakeholders are helped to make more intelligent decisions. 本发明公开了一种可解释的交通流预测堆叠方法，包括如下步骤：S1、收集交通流量预测相关时空特征及数据预处理；S2、搭建堆叠模型、调试与检验；S3、通过一致性聚类思想评估各个时空特征对预测结果的影响。本发明在提升预测结果的准确度的同时，保证了预测结果的可解释性；提供了一个旨在增强预测精度和模型可解释性的新框架，从而帮助交通专业人员和利益相关者做出更明智的决策。