Predicting network flows from speeds using open data and transfer learning

Abstract

Traffic flow/volume data are commonly used to calibrate and validate traffic simulation models. However, these data are generally obtained from stationary sensors (e.g. loop detectors), which are expensive to install and maintain and cover a small number of locations in the transport network. On the other hand, Floating Car Data (FCD) are readily available at the network level, usually from a sample of vehicles. We present an indirect traffic flow estimation approach using transfer learning to address the traffic flow data scarcity and model generalization across cities. Using two cities (Paris and Madrid) as study areas, we demonstrate the indirect estimation using only exogenous features for flow prediction, mirroring limited predictive features without past link flows. Subsequently, we use the model pre‐trained on data from Paris city and test on data from Madrid city, and investigate the scenarios for successful transfer learning. Overall, the training set must adequately capture the flow‐speed relationship for successful indirect flow estimation. Transfer learning is beneficial when the data for the target task is minimal, in which case transferred models outperform newly trained models from scratch. Using real‐world and publicly available data, our approach and models can help scale a smaller traffic flow dataset to a larger sample across cities.

Scaling traffic variables from sensors sample to the entire city at high spatiotemporal resolution with machine learning: applications to the Paris megacity

“…Air Parif, the regional air quality agency also uses sensor data to calibrate [30] an emission model for monitoring air quality in the area. Combined with Uber speed data [31], this dataset was also employed by Mahajan et al [32] who used a transfer learning method to predict the road flow from Paris data in the city of Madrid in 2019. The Uber speed data is no longer available since October 2023.…”

“…Thus, the geometry characteristics were matched with an independent road geospatial database to obtain the characteristics of each road link. As in Mahajan et al [32], we matched the geometries to OpenStreetMap [29] data using PyTrack [38] rather than SharedStreets [39] which seems to be no longer maintained. This Python toolkit uses Hidden Markov Model [40] to detect the most probable path of points retrieved from the Paris open data geometry through the OSM network.…”

“…We built a model of target labels Q and O using all the predictors of Table 1. We employed a random forest regressor and tested other models like gradient boosting used as a baseline for comparison in other studies [32]. Random forest regression is a widely-used technique that deals well with correlated features and big datasets.…”

Overcoming the Barriers of Using Linked Open Data in Smart City Applications

“…Also, in the European Open Data Ecosystem emerged initiatives such as the EU Open Data Portal and the European Data Portal, in the Data Spaces Ecosystem emerged initiatives to share data in the business ecosystem. The most relevant European initiatives are International Data Spaces (IDS) 17 and Gaia-X 18 . They define a common framework to establish federated data spaces that preserve the data sovereignty of each participant.…”

“…They have also been used to develop solutions in many other fields such as digital twins [12], cultural heritage [13], agriculture [14], medicine [15], etc. Even OD has been used in artificial intelligence by providing datasets to train machine learning models [16], [17].…”