Next Point of Interest (POI) recommendation primarily predicts future activities based on users’ past check-in data and current status, providing significant value to users and service providers. We observed that the popular check-in times for different POI categories vary. For example, coffee shops are crowded in the afternoon because people like to have coffee to refresh after meals, while bars are busy late at night. However, existing methods rarely explore the relationship between POI categories and time, which may result in the model being unable to fully learn users’ tendencies to visit certain POI categories at different times. Additionally, existing methods for modeling time information often convert it into time embeddings or calculate the time interval and incorporate it into the model, making it difficult to capture the continuity of time. Finally, during POI prediction, various weighting information is often ignored, such as the popularity of each POI, the transition relationships between POIs, and the distances between POIs, leading to suboptimal performance. To address these issues, this article proposes a novel next POI recommendation framework called Graph Disentangler with POI Weighted Module (GDPW) . This framework aims to jointly consider POI category information and multiple POI weighting factors. Specifically, the proposed GDPW learns category and time representations through the Global Category Graph and the Global Category-Time Graph. Then, we disentangle category and time information through contrastive learning. After prediction, the final POI recommendation for users is obtained by weighting the prediction results based on the transition weights and distance relationships between POIs. We conducted experiments on two real-world datasets, and the results demonstrate that the proposed GDPW outperforms other existing models, improving performance by 3% to 11%.

Predicting human mobility across multiple cities is essential for urban applications but remains challenging due to the complex and diverse spatiotemporal dynamics in human trajectories. While recent work often leverages language modeling by treating trajectories as sequences for next-location prediction, these approaches typically rely on raw movement data, process long trajectories without distinguishing between individual trips, and use a single model for all users within a city. To address these limitations, this paper presents MoVER, a transformer encoder-decoder that enriches trajectory representations with location profiles and explicit trip separators. Furthermore, we introduce a clustering-based finetuning strategy to handle user heterogeneity by tailoring models to user groups of similar travel patterns. MoVER outperforms baselines on a validation set and achieves a top-7 ranking among over 50 participating teams in the 14th SIGSPATIAL Cup competition (GISCUP 2025). The code is publicly available at: https://github.com/knowledge-computing/giscup2025-mover.

Historical geologic maps contain rich geospatial information—such as rock units, faults, folds, and bedding planes—that is critical for assessing mineral resources essential to renewable energy, electric vehicles, and national security. However, digitizing maps remains a labor-intensive and time-consuming task. We present DIGMAPPER, a modular, scalable system developed in collaboration with the United States Geological Survey (USGS) to automate the digitization of geologic maps. DIGMAPPER features a fully dockerized, workflow-orchestrated architecture that integrates state-of-the-art deep learning models for map layout analysis, feature extraction, and georeferencing. To overcome challenges such as limited training data and complex visual content, our system employs innovative techniques, including in-context learning with large language models, synthetic data generation, and transformer-based models. Evaluations on over 100 annotated maps from the DARPA-USGS dataset demonstrate high accuracy across polygon, line, and point feature extraction, and reliable georeferencing performance. Deployed at USGS, DIGMAPPER significantly accelerates the creation of analysis-ready geospatial datasets, supporting national-scale critical mineral assessments and broader geoscientific applications.

Historical maps often suffer from coloring errors caused by artifacts during map production or scanning. These errors result in color mismatches between important map features (e.g., polygon layers) and their corresponding map keys, which hinders both human interpretation and automated feature extraction. This paper targets the problem of automatically correcting polygon coloring errors in historical maps using only in-map information, such as the map keys. The challenge lies in the diverse visual representations of map keys and variations in coloring errors, which differ significantly both within and across maps. We propose a machine-learning model that automatically identifies and corrects color inconsistencies between map polygon layers and their visual appearances defined by the map keys on the same map. Our approach leverages polygon metadata, such as map keys describing the visual and semantic properties of each polygon on maps, to detect mismatches in color histograms and representations and recolor the incorrect areas in the map content accordingly. We evaluate our approach on USGS geological maps; it outperforms comparative methods by at least 7.5%. In addition, our approach improves the downstream automated polygon-extraction task by 18.0% in precision.

Black-and-white draft geological maps, produced during fieldwork, often contain dense handwritten annotations overlaid on monochromatic contour basemaps. Although interpretable in grayscale, the lack of color makes it difficult to visually distinguish overlapping or adjacent geological units, especially when boundaries are unclear and annotation styles vary. However, colorizing these draft maps is labor-intensive but essential, as they may be the only source of detailed geological information for certain regions. This hinders both human interpretation and downstream tasks such as map digitization and critical mineral resource assessment. We target the problem of automated colorization of draft geological maps. The challenge lies in interpreting noisy visual cues from uncolored sketches and assigning appropriate colors according to their geological categories. We propose a novel machine learning approach that exploits polygon metadata, including map keys that explicitly define geological units and implicitly suggest their intended colors, along with the semantic interpretation of the sketch content in the maps, to assign colors to the draft maps accordingly. We evaluate our method on USGS draft geological maps; it outperforms comparative methods by 15.7%. In addition, our approach improves downstream polygon-extraction performance by 9% in F1 score.

Understanding the semantic structure of tabular data is essential for data integration and discovery. Specifically, the goal is to annotate columns in a tabular source with types and relationships between them using classes and predicates of a target ontology. Previous work either requires trained labeled data or exploits the overlapping data between the table data and a knowledge graph to predict types and relationships. However, these approaches cannot be used in a new domain with limited labeled data. To address this issue, we propose a novel domain-independent approach to estimate a score reflecting the semantic relatedness between a table column and an ontology class or property using the table metadata and data. Our empirical evaluation demonstrates that our approach significantly outperforms strong baselines based on large language models.

With the rapid development of electric vehicles that facilitate sustainable transportation, the need to ensure their accessibility has increased. Determining the optimal locations for extending an existing charging station network is crucial to addressing rising infrastructure needs and reducing environmental impacts. The 13th ACM SIGSPATIAL Cup competition (GISCUP 2024) targets the problem of optimal minimum charging station deployment within an existing network. The recommendation of station locations aims to maximize accessibility over long-distance travel and ensure both coverage and proximity to points of interest or areas with high vehicle usage, while minimizing traffic congestion and reducing impacts on the power grid and the existing charging network. We leverage the categorical Voronoi diagram with hierarchical reconciliation to recommend stations that dovetail with the existing charging network under zero access to historical data. Our approach maximizes station deployment in high-demand areas with equitable distribution. Meanwhile, it complements the existing network to minimize unfavorable impacts on the power grid. Qualitative evaluation on a large-scale real-world dataset shows that our approach effectively balances objectives across areas with diverse geographical and demographic characteristics. In addition, our approach ranked in the top five among all participants in the GISCUP 2024.

Road inspection and cleaning are crucial to securing driving safety. Deploying a fleet of robots that run through a city can inspect and clean pavements without causing road closure. To achieve high coverage, one has to prevent robots from going through a road segment more than once. However, robots may need more than one visit to a particular road segment to inspect a defect. The uncertain success rate of defect inspection and the unknown maximum number of defects hinder the efficacy. Such uncertainty and constraints in objectives can also be seen in security patrolling, trip planning, and network maintenance. We target the problem of multi-agent scheduling with bounded objectives. The scheduling aims for maximum road network coverage while ensuring sufficient visits to particular road segments for defect identification of an uncertain subject, such as potholes and faded markings during road inspection or crimes and parking violations during security patrolling. We leverage an approximate bi-objective algorithm and propose a hierarchical circular route-planning algorithm. Our approach maximizes the road coverage among robots and decreases the search space when maximizing defect identification. Evaluation on a real-world dataset shows that our approach achieves the Pareto optimal among comparative methods, outperforming existing methods by at least one optimization objective.

Private vehicles are a direct means to bring people from one place to their desired destinations. However, no omniscient dispatcher is handling the origin-destination of vehicles and the availability of stationary resources, such as parking spaces or charging stations. Competitive cruising for stationary resources leads to environmental pollution and is a waste of drivers' time. We focus on the problem of distributed stationary-resource searching with predefined destinations under a multi-agent scenario. It is a distributed route planning problem with global optimization objectives. We present a probabilistic approach to achieving indirect resource coordination and latent agent cooperation in a distributed manner. Our approach treats the estimated availability of stationary resources as a reference and guides each agent based on their preferences. We evaluate our approach on four real-world datasets. Our approach outperforms state-of-the-art methods by 5% in multi-criteria optimization.

Locating undiscovered deposits of critical minerals requires accurate geological data. However, most of the 100,000 historical geological maps of the United States Geological Survey (USGS) are in raster format. This hinders critical mineral assessment. We target the problem of extracting geological features represented as polygons from raster maps. We exploit the polygon metadata that provides information on the geological features, such as the map keys indicating how the polygon features are represented, to extract the features. We present a metadata-driven machine-learning approach that encodes the raster map and map key into a series of bitmaps and uses a convolutional model to learn to recognize the polygon features. We evaluated our approach on USGS geological maps; our approach achieves a median F1 score of 0.809 and outperforms state-of-the-art methods by 4.52%.

Hybrid computational intelligent systems that synergize learning-based inference models and route planning strategies have thrived in recent years. In this article, we focus on the non-monotonicity originated from heterogeneous urban data, as well as heuristics based on neural networks, and thereafter formulate the traveling transporter problem (TTP). TTP is a multi-criteria optimization problem and may be applied to the circular route deployment in public transportation. In particular, TTP aims to find an optimized route that maximizes passenger flow according to a neural-network-based inference model and minimizes the length of the route given several constraints, including must-visit stations and the requirement for additional ones. As a variation of the traveling salesman problem (TSP), we propose a framework that first recommends new stations’ location while considering the herding effect between stations, and thereafter combines state-of-the-art TSP solvers and a metaheuristic named Trembling Hand , which is inspired by self-efficacy for solving TTP. Precisely, the proposed Trembling Hand enhances the spatial exploration considering the structural patterns, previous actions, and aging factors. Evaluation conducted on two real-world mass transit systems, Tainan and Chicago, shows that the proposed framework can outperform other state-of-the-art methods by securing the Pareto-optimal toward the objectives of TTP among comparative methods under various constrained settings.

In this work, a novel decision assistant system for urban transportation, called Route Scheme Assistant (RSA), is proposed to address two crucial issues that few former researches have focused on: route-based passenger flow (PF) inference and multivariant high-PF route recommendation. First, RSA can estimate the PF of arbitrary user-designated routes effectively by utilizing Deep Neural Network (DNN) for regression based on geographical information and spatial-temporal urban informatics. Second, our proposed Bidirectional Prioritized Spanning Tree (BDPST) intelligently combines the parallel computing concept and Gaussian mixture model (GMM) for route recommendation under users’ constraints running in a timely manner. We did experiments on bus-ticket data of Tainan and Chicago and the experimental results show that the PF inference model outperforms baseline and comparative methods from 41% to 57%. Moreover, the proposed BDPST algorithm's performance is not far away from the optimal PF and outperforms other comparative methods from 39% to 71% in large-scale route recommendations.

Transportation between satellite cities or inside the city center has always been a crucial factor in contributing to a better quality of life. This paper focuses on a multi-criteria distributed competitive route planning for parking slot cruising in regions where neither real-time nor historical availability of parking slots is accessible. An inference-than-planning framework is proposed for solving the parking slot searching using a zero-information distributed model with an availability inference for parking slots in areas with no sensor coverage. Meanwhile, a proposed Conntrans algorithm is suggested as a two-stage structure with three relaxing policies: adjacent cruising, on-orbital annealing, and orbital transitioning. The evaluation is conducted based on the simulation in a publicly accessible real-world parking data from SFPark in San Francisco; the area is divided into 3 separated regions with different urban characteristics. Overall results show that the proposed availability inference model can retrieve decent performance. Furthermore, Conntrans is able to outperform baselines and state-of-the-arts in overall score by at most 77% with a success rate at around 97% and maintains the quality of solutions under various circumstances.

Traffic-related forecasting plays a critical role in determining transportation policy, unlike traditional approaches, which can only make decisions based on statistical results or historical experience. Through machine learning, we are able to capture the potential interactions between urban dynamics and find their mutual interactions in a spatial context. However, despite a plethora of traffic-related studies, few works have explored predicting the impact of congestion. Therefore, this paper focuses on predicting how a car accident leads to traffic congestion, especially the length of time it takes for the congestion to occur. Accordingly, we propose a novel model named Dual-Attention Multi-Scale Graph Convolutional Networks (DAMGNet) to address this issue. In this proposed model, heterogeneous data such as accident information, urban dynamics, and various highway network characteristics are considered and combined. Next, the context encoder encodes the accident data, and the spatial encoder captures the hidden features between multi-scale Graph Convolutional Networks (GCNs). With our designed dual attention mechanism, the DAMGNet model is able to effectively learn the correlation between features. The evaluations conducted on a real-world dataset prove that our DAMGNet has a significant improvement in RMSE and MAE over other comparative methods.

Multi-criteria path planning is an important combinatorial optimization problem with broad real-world applications. Finding the Pareto-optimal set of paths ideal for all requiring features is time-consuming and unclear to obtain the subset of optimal paths efficiently for multiple origin states in the planning space. Meanwhile, due to the rise of deep learning, hybrid systems of computational intelligence thrive in recent years. When facing non-monotonic data or heuristics derived from pre-trained neural networks, most of the existing methods for the one-to-all path problem fail to find an ideal solution. We employ Gaussian mixture model to propose a target-prioritized searching algorithm called Multi-Source Bidirectional Gaussian-Prioritized Spanning Tree (BiasSpan) in solving this non-monotonic multi-criteria route planning problem given constraints including range, must-visit vertices, and the number of recommended vertices. Experimental results on mass transportation system in Tainan and Chicago cities show that BiasSpan outperforms comparative methods from 7% to 24%and runs in a reasonable time compared to state-of-art route-planning algorithms.

Urban areas need to deploy a lot of services and stations. This work considers the issue of establishing new branches for a certain service. Given a number of stations we plan to construct, our goal is to recommend locations as deploy placements and transportation cost could be efficiently reduced by jointly considering road network, existing stations and spatial event data. Our model can be divided into four parts: 1) Adopting DBSCAN clustering method to find hot spots of spatial events. 2) Doing community detection for road network to split the road network to smaller components. 3) Exploiting a refined closeness centrality to identify a good candidate location in each community. 4) Developing a greedy-based distance minimized method to establish stations sequentially. The results show our solution is effective and efficient for a large crime event dataset of Chicago.

In this work, we propose a novel system, called Route Planning Maker (RPM) to help the government or transportation companies to design new route services in the city. The RPM system has a flexible user interface that allows users design the nearby areas of a new route and further deploying new stations. Moreover, based on user-designed arbitrary transportation routes and the expected locations of stations, the RPM system provides an intelligent function to infer passenger flows in certain time intervals so that the user can estimate the effectiveness of designed routes. To capture the spatial-temporal factors correlated with passenger flows, we propose to combine dynamic features such as human mobility, passenger volume of existing routes, and static features, including road network structure, point-of-interests (POI), station placement of existing routes and local population structure. Finally, to combine these features, we modified Deep Neural Network (DNN) for regression to derive the passenger flow for each given designated route. The experiments on the Tainan's bus-ticket data outperform baseline methods for 75%.