Introduction to the Special Issue on Machine Learning and Location Data

Introduction to the Special Issue on Machine Learning and Location DataDue to the rapid development of location acquisition technologies, big location data are being collected from various data sources such as connected cars, sharing bikes, smartphones, sensors, social media, and Wi-Fi access points.This led to the proposal of various spatial algorithms, models, and systems to understand and model these data for intelligent transportation, business intelligence, public health, social economics, urban planning, urban resilience, and environmental sustainability.This ACM TSAS special issue on Machine Learning and Location Data includes 14 articles introducing novel and innovative techniques applied to location data.We received 36 submissions in total, and 14 of 36 were accepted for publication after at least one revision cycle.These 14 articles are organized and published in two separate issues.In this issue, we have 7 exciting articles covering the fundamental issues in location data-those include the intricacies of the underlying spatio-temporal network representations, the dynamic correlation of spatio-temporal data, including the sequential and causal behaviors, and, finally, the anomalies due to location dynamics during extreme events.The first 2 articles in this issue deal with the spatio-temporal characteristics of the road network, the complexity and the varying resolutions of the network.The first article, titled "Spatio-temporal Dual Graph Neural Networks for Travel Time Estimation, " presents a graph-based model that incorporates node-and edge-wise graphs to characterize adjacency of intersections and capture the network architecture at multiple scale to solve travel time estimation problem.The second article, titled "Deformation Gated Recurrent Network for Lane-level Abnormal Driving Behavior Recognition, " presents conditional random field model to detect anomalous behavior at lane-level granularity, rather than at road-level.The next 2 articles tackle the complex spatio-temporal correlations in location data.The third article, titled "Adaptive Spatio-temporal Graph Learning for Bus Station Profiling, " aims to model the interactions between spatio-temporal correlations, capture the shift of distributions, and model the long-term patterns of bus station profiles.The fourth article, titled "Adaptive Joint Spatiotemporal Graph Learning Network for Traffic Data Forecasting, " aims at modeling the dynamic spatio-temporal correlations in long-term intervals towards more effective traffic forecasting.The next 2 articles focus on a set of sequential prediction and recommendation tasks.In the fifth article, titled "Geography-aware Heterogeneous Graph Contrastive Learning for Travel Recommendation, " a multi-view contrastive learning method is proposed to fuse the heterogeneous information from the user and the travel product representation to handle sparsity and provide more diverse views.In the sixth article, titled "Causal Probabilistic Spatio-temporal