A common need for artificial intelligence models in the broader geoscience is to encode various types of spatial data, such as points, polylines, polygons, graphs, or rasters, in a hidden embedding space so that they can be readily incorporated into deep learning models. One fundamental step is to encode a single point location into an embedding space, such that this embedding is learning-friendly for downstream machine learning models. We call this process location encoding. However, there lacks a systematic review on location encoding, its potential applications, and key challenges that need to be addressed. This paper aims to fill this gap. We first provide a formal definition of location encoding, and discuss the necessity of it for GeoAI research. Next, we provide a comprehensive survey about the current landscape of location encoding research. We classify location encoding models into different categories based on their inputs and encoding methods, and compare them based on whether they are parametric, multi-scale, distance preserving, and direction aware. We demonstrate that existing location encoders can be unified under one formulation framework. We also discuss the application of location encoding. Finally, we point out several challenges that need to be solved in the future.
Good characterization of traffic interactions among urban roads can facilitate traffic-related applications, such as traffic control and short-term forecasting. Most studies measure the traffic interaction between two roads by their topological distance or the correlation between their traffic variables. However, the distance-based methods neglect the spatial heterogeneity of roads’ traffic interactions, while the correlation-based methods cannot capture the non-linear dependency between two roads’ traffic variables. In this paper, we propose a novel approach called Road2Vec to quantify the implicit traffic interactions among roads based on large-scale taxi operating route data using a Word2Vec model from the natural language processing (NLP) field. First, the analogy between transportation elements (i.e., road segment, travel route) and NLP terms (i.e., word, document) is established. Second, the real-valued vectors for road segments are trained from massive travel routes using the Word2Vec model. Third, the traffic interaction between any pair of roads is measured by the cosine similarity of their vectors. A case study on short-term traffic forecasting is conducted with artificial neural network (ANN) and support vector machine (SVM) algorithms to validate the advantages of the presented method. The results show that the forecasting achieves a higher accuracy with the support of the Road2Vec method than with the topological distance and traffic correlation based methods. We argue that the Road2Vec method can be effectively utilized for quantifying complex traffic interactions among roads and capturing underlying heterogeneous and non-linear properties.