Public transportation plays a vital role in supporting sustainable, accessible, environment-friendly, and equitable urban mobility. However, challenges such as poor first- and last-mile connectivity, limited service coverage, and inefficient use of space continue to limit its effectiveness and uptake. Autonomous vehicles (AVs) offer new opportunities to address these limitations by enhancing flexibility, improving access, and complementing existing transit systems. This vision paper explores how a diverse fleet of AVs, including cars, shuttles, pods, scooters, and buses, can be integrated into public transport to form an adaptive, multimodal, and data-driven mobility ecosystem. We outline key research directions spanning fleet coordination, spatial deployment, infrastructure planning, and intelligent transportation platforms. We highlight the need for interdisciplinary research at the intersection of spatial computing, transportation systems, artificial intelligence, and urban data infrastructure. Our aim is to inform and inspire future efforts toward building autonomous mobility systems that are efficient, inclusive, and future-ready.

Recent advancements in foundation models have improved autonomous tool usage and reasoning, but their capabilities in map-based reasoning remain underexplored. To address this, we introduce MapEval, a benchmark designed to assess foundation models across three distinct tasks—textual, API-based, and visual reasoning—through 700 multiple-choice questions spanning 180 cities and 54 countries, covering spatial relationships, navigation, travel planning, and real-world map interactions. Unlike prior benchmarks that focus on simple location queries, MapEval requires models to handle long-context reasoning, API interactions and visual map analysis, making it the most comprehensive evaluation framework for geospatial AI. On evaluation of 30 foundation models, including Claude-3.5-Sonnet, GPT-4o, Gemini-1.5-Pro, none surpasses 67% accuracy, with open-source models performing significantly worse and all models lagging over 20% behind human performance. These results expose critical gaps in spatial inference, as models struggle with distances, directions, route planning, and place-specific reasoning, highlighting the need for better geospatial AI to bridge the gap between foundation models and real-world navigation.

Code synthesis, which requires a deep understanding of complex natural language (NL) problem descriptions, generation of code instructions for complex algorithms and data structures, and the successful execution of comprehensive unit tests, presents a significant challenge.Thus, while large language models (LLMs) demonstrate impressive proficiency in natural language processing (NLP), their performance in code generation tasks remains limited.In this paper, we introduce a new approach to code generation tasks leveraging the multi-agent prompting that uniquely replicates the full cycle of program synthesis as observed in human developers.Our framework, MapCoder, consists of four LLM agents specifically designed to emulate the stages of this cycle: recalling relevant examples, planning, code generation, and debugging.After conducting thorough experiments, with multiple LLMs ablations and analyses across eight challenging competitive problem-solving and program synthesis benchmarks-MapCoder showcases remarkable code generation capabilities, achieving their new state-of-the-art (pass@1) results-(HumanEval 93.9%, MBPP 83.1%, APPS 22.0%, CodeContests 28.5%, and xCodeEval 45.3%).Moreover, our method consistently delivers superior performance across various programming languages and varying problem difficulties.

While large multimodal models (LMMs) have obtained strong performance on many multimodal tasks, they may still hallucinate while generating text.Their performance on detecting salient features from visual data is also unclear.In this paper, we develop a framework to generate faithful and salient text from mixed-modal data, which includes images and structured data ( represented in knowledge graphs or tables).Specifically, we train a small vision critic model to identify hallucinated and non-salient features from the image modality.The critic model also generates a list of salient image features.This information is used in the post editing step to improve the generation quality.Experiments on two datasets show that our framework improves LMMs' generation quality on both faithfulness and saliency, outperforming recent techniques aimed at reducing hallucination.The dataset and code are available at https://github.com/TahsinaHashem/FaithD2T. Prompt Template New for LMM House Dataset Test Sample-1 (Linearized Graph with Images) #INPUT Prompt: Consider the uploaded Image of a House List the Key features of this image for House Advertising. #OUTPUT: Key features of image:1:The key features of this image are: a large glass wall, a wooden deck with chairs, a swimming pool, and a view of the surrounding city. Key features of image:2:This image shows a rooftop patio with a large table and chairs, a small table with two chairs, and a large planter with plants.The view from the rooftop is of a cityscape with skyscrapers and trees in the foreground.The sky is clear and sunny, and there is a blue and white striped awning over the patio.The patio is surrounded by a railing made of metal and glass, and there are several large windows on the side of the building.The building is made of concrete and glass, with a modern design. Key features of image:3The image shows a modern, minimalist bathroom with a white wall tiles, a grey toilet, and a mirror above the sink.The toilet is mounted on the wall and there is a white towel rack next to it.The sink is made of white porcelain and there are two towels hanging from the rack.The shower area is visible in the background, and there is a grey tiled floor.The walls have large white tiles and there is a small window with a white curtain.The overall impression is of a clean, modern, and functional bathroom.………………….. …………………..Figure 3: Prompt Template for LMM to generate key features of the image for House dataset Prompt Template New for LMM House Dataset Test Sample-1 (Linearized Graph with Images) #INPUT Prompt: Consider the House Linearized Graph: This house location Manningham.This house property type Apartment.This house address 642/

Seismic intensity prediction from early or initial seismic waves received by a few seismic stations can enhance Earthquake Early Warning (EEW) systems, particularly in ground motion-based approaches like PLUM. While many operational EEW systems currently utilize point-source-based models that estimate the warning area based on magnitude and distance measures, direct intensity prediction offers a potential improvement in accuracy and reliability. In this paper, we propose a novel deep learning approach, Seismic Contrastive Graph Neural Network (SC-GNN), for highly accurate seismic intensity prediction using a small portion of initial seismic waveforms from a few seismic stations. The SC-GNN consists of two key components: (i) a graph neural network (GNN) to propagate spatiotemporal information through a graph-like structure representing seismic station distribution and wave propagation, and (ii) a self-supervised contrastive learning component to train the network with larger time windows and enable predictions using shorter initial waveforms. The efficacy of our approach is demonstrated through experiments on three real-world seismic datasets, where it shows superior performance over existing techniques, including a significant reduction in mean squared error (MSE) and the lowest standard deviation of error, indicating its robustness, reliability, and strong positive relationship between predicted and actual values. Notably, the SC-GNN model maintains superior performance even with 5s input waveforms, making it especially suitable for enhancing EEW applications.

Alternative route planning requires finding$k$alternative paths (including the shortest path) between a given source and target. These paths should be significantly different from each other and meaningful/natural (e.g., must not contain loops or unnecessary detours). While there exists many work on finding high-quality alternative paths, the existing techniques are computationally expensive and are unable to accommodate the high volume of queries required by modern navigation systems. To address this, in this paper, we propose an efficient approach to compute high-quality alternative paths. Our approach employs hub-labeling to efficiently identify candidate alternative paths. The candidate paths are then ranked considering multiple quality metrics and the top-$k$alternative paths are returned. We propose several non-trivial optimizations to significantly improve the computation time. In our experimental study, we conduct experiments on three diverse real-world road networks and compare our proposed algorithm against six state-of-the-art algorithms. The results demonstrate that our algorithm is not only up to 3 orders of magnitude faster compared to most algorithms but also consistently generates alternative paths that are comparable or even superior in terms of quality across various metrics.

The advancement of indoor location-aware technologies enables a wide range of location based services in indoor spaces. In this paper, we formulate a novel Indoor Facility Location Selection (IFLS) query that finds the optimal location for placing a new facility (e.g., a coffee station) in an indoor venue (e.g., a university building) such that the maximum distance of all clients (e.g., staffs/students) to their nearest facility is minimized. To the best of our knowledge we are the first to address this problem in an indoor setting. We first adapt the state-of-the-art solution in road networks for indoor settings, which exposes the limitations of existing approaches to solve our problem in an indoor space. Therefore, we propose an efficient approach which prunes the search space in terms of the number of clients considered, and the total number of facilities retrieved from the database, thus reducing the total number of indoor distance calculations required. The key idea of our approach is to use a single pass on a state-of-the-art index for an indoor space, and reuse the nearest neighbor computation of clients to prune irrelevant facilities and clients. We evaluate the performance of both approaches on four indoor datasets. Our approach achieves a speedup from 2.84× to 71.29× for synthetic data and 97.74× for real data over the baseline.

Knowledge Graph (KG)-to-Text generation aims at generating fluent natural-language text that accurately represents the information of a given knowledge graph. While significant progress has been made in this task by exploiting the power of pre-trained language models (PLMs) with appropriate graph structure-aware modules, existing models still fall short of generating faithful text, especially when the ground-truth natural-language text contains additional information that is not present in the graph. In this paper, we develop a KG-to-text generation model that can generate faithful natural-language text from a given graph, in the presence of noisy reference text. Our framework incorporates two core ideas: Firstly, we utilize contrastive learning to enhance the model's ability to differentiate between faithful and hallucinated information in the text, thereby encouraging the decoder to generate text that aligns with the input graph. Secondly, we empower the decoder to control the level of hallucination in the generated text by employing a controllable text generation technique. We evaluate our model's performance through the standard quantitative metrics as well as a ChatGPT-based quantitative and qualitative analysis. Our evaluation demonstrates the superior performance of our model over state-ofthe-art KG-to-text models on faithfulness.

With the advent of location-based social networks, users can tag their daily activities in different locations through check-ins. These check-in locations signify user preferences for various socio-spatial activities and can be used to improve the quality of services in some applications such as recommendation systems, advertising, and group formation. To support such applications, in this paper, we formulate a new problem of identifying top-k Socio-Spatial co-engaged Location Selection (SSLS) for users in a social graph, that selects the best set of k locations from a large number of location candidates relating to the user and her friends. The selected locations should be (i) spatially and socially relevant to the user and her friends, and (ii) diversified both spatially and socially to maximize the coverage of friends in the socio-spatial space. To address the NP-hard and challenging problem, we first develop an exact solution by designing some pruning strategies, and also develop an approximate solution by deriving relaxed bounds and advanced termination rules. To accelerate the efficiency, we further develop a fast exact approach and a meta-heuristic approximate approach. Finally, extensive experiments are conducted to evaluate the performance of our proposed algorithms against three adapted existing methods using four real-world datasets.

Trip itinerary recommendation finds an ordered sequence of Points-of-Interest (POIs) from a large number of candidate POIs in a city. In this paper, we propose a deep learning-based framework, called DeepAltTrip, that learns to recommend top-$k$alternative itineraries for given source and destination POIs. These alternative itineraries would be not only popular given the historical routes adopted by past users but also dissimilar (or diverse) to each other. The DeepAltTrip consists of two major components: (i)Itinerary Net(ITRNet) which estimates the likelihood of POIs on an itinerary by using graph autoencoders and two (forward and backward) LSTMs; and (ii) a route generation procedure to generate$k$diverse itineraries passing through relevant POIs obtained using ITRNet. For the route generation step, we propose a novel sampling algorithm that can seamlessly handle a wide variety of user-defined constraints. To the best of our knowledge, this is the first work thatlearnsfrom historical trips to provide a set of alternative itineraries to the users. Extensive experiments conducted on eight popular real-world datasets show the effectiveness and efficacy of our approach over state-of-the-art methods.

Abstract Alternative pathfinding requires finding a set of k alternative paths (including the shortest path) between a given source s and a target t. Intuitively, these paths should be significantly different from each other and meaningful/natural (e.g., must not contain loops or unnecessary detours). While finding alternative paths in road networks has been extensively studied, to the best of our knowledge, we are the first to formally study alternative pathfinding in game maps which are typically represented as Euclidean planes containing polygonal obstacles. First, we adapt the existing techniques designed for road networks to find alternative paths in the game maps. Then, based on our web-based system that visualises alternative paths generated by different approaches, we conduct a user study that shows that the existing road network approaches generate high-quality alternative paths when adapted for the game maps. However, these existing approaches are computationally inefficient especially when compared to the state-of-the-art shortest path algorithms. Motivated by this, we propose novel data structures and exploit these to develop an efficient algorithm to compute high-quality alternative paths. that shows that the existing road network approaches generate high-quality alternative paths in game maps. Our extensive experimental study demonstrates that our proposed algorithm is more than an order of magnitude faster than the existing approaches and returns alternative paths of comparable quality. Furthermore, our algorithm is comparable to a state-of-the-art shortest path algorithm in terms of running time.

In this COVID-19 pandemic era, students with Autism Spectrum Disorder (ASD) are struggling to adapt to classes in the online environment using Google Meet or Zoom. Failing to keep sustained attention in the class is a common problem for students with ASD. In face-to-face classes, teachers can track a student's behavior and activity to infer the student's attentiveness level and act accordingly. However, it becomes difficult for a teacher to monitor the attentiveness level of multiple students simultaneously on online platforms like Zoom. Detecting the attentiveness level of a student and notifying the teacher in an automated way can play a crucial role in improving the learning outcome. In this paper, we propose the first deep learning based attentiveness level prediction technique for students with ASD. Our model detects the behavior (e.g., unusual movement, gaze etc.) and activities from real-time videos and uses them as features to classify the attentiveness level as low, mid and high. Existing state-of-the-art techniques to detect the attentiveness level of typically developed students using gaze or facial expression cannot be trivially extended for students with ASD as they do not exhibit regular and consistent behavior. We collect video data belonging to different classes covering various types of activities over a long period, train our classifier, and run extensive experiments to validate the prediction performance. Our solution outperforms existing baselines by a large margin.

Anger detection from conversations has a number of real-life applications that include improving interpersonal communications, providing customer services, and enhancing work-place performance. In this paper, we propose novel deep learning-based approaches for both offline and online anger detection from audio-textual data obtained from real-life conversations. For offline anger detection, which detects the anger of a given audio-textual conversation, we introduce an ensemble approach that adapts attention-based CNN architecture, gender classifier, and BERT-based textual features to derive the anger of a conversion. On the other hand, for online anger detection, which predicts anger in the conversation of the subsequent timestamps from the conversations of previous timestamps, we propose a transformer-based audio and textual ensemble technique to predict the anger of a future conversation. We demonstrate the efficacy of our proposed approaches using two datasets: the Bengali call-center dataset and the IEMOCAP dataset. Experimental results show that our proposed approaches outperform the state-of-the-art baselines by a significant margin. For offline anger, our model achieves an F1 score of 85.5% on the Bengali dataset and 91.4% on the IEMOCAP dataset. For online anger, our model yields an F1 score of 66.9% on the Bengali dataset and 67.7% on the IEMOCAP dataset.

Computing multiple alternative routes from a source$s$to a target$t$has received significant research attention. However, it is unclear which of the existing approaches generates alternative routes of better quality because the quality of these alternatives is mostly subjective. Motivated by this, in this paper, we present a user study conducted on the road networks of Melbourne, Dhaka and Copenhagen comparing four of the most popular existing approaches including Google Maps. We report the average ratings received by the four approaches, and our statistical analysis shows that there is no credible evidence that the four approaches receive different ratings on average. We also discuss the limitations of this user study and recommend the readers interpret these results with caution.

Many modern navigation systems and map-based services do not only provide the fastest route from a source location$s$to a target location$t$but also provide a few alternative routes to the users as more options to choose from. Consequently, computing alternative paths has received significant research attention. However, it is unclear which of the existing approaches generates alternative routes of better quality because the quality of these alternatives is mostly subjective. Motivated by this, in this paper, we present a user study conducted on the road networks of Melbourne, Dhaka and Copenhagen that compares the quality (as perceived by the users) of the alternative routes generated by four of the most popular existing approaches including the routes provided by Google Maps. We also present a web-based demo system that can be accessed using any internet-enabled device and allows users to see the alternative routes generated by the four approaches for any pair of selected source and target. We report the average ratings received by the four approaches and our statistical analysis shows that there is no credible evidence that the four approaches receive different ratings on average. We also discuss the limitations of this user study and recommend the readers to interpret these results with caution because certain factors may have affected the participants’ ratings.

Traditional face-to-face health consultation-based systems largely failed to attract teenagers to get reproductive and sexual health supports from doctors and practitioners in Bangladesh as ‘sex’ or ‘adolescent’ related issues are considered social taboos and are rarely discussed openly with anyone. This has damaging implications for the physiological and mental well-being of a large group of people. In this paper, we study chatbot’s effectiveness to assist adolescents in seeking reproductive and sexual health supports by analyzing the responses from 256 participants, including adolescents and medical personnel from six different regions of Bangladesh. We prototype an interactive chatbot, namely AdolescentBot, and analyzed users’ communication patterns, feelings, and contexts of use as the first point of support for getting adolescence related health advice. Our analysis finds that a chatbot can satisfy most of the users’ queries, and the majority of the queries are associated with wrong-beliefs. Finally, we discuss ethical and societal issues with chatbot usage and recommend a set of design propositions for the AdolescentBot.

In this paper, we propose an iterative approach to estimate the frequent itemsets with high accuracy while satisfying the local differential privacy (LDP). The key component behind the improved accuracy of the estimated frequent itemsets by our approach is our novel two-level randomization technique for guaranteeing the LDP. Our randomization technique exploits the correlation of the presence of items in a user’s itemset, which has not been considered before. We present a mathematical proof that shows that our approach satisfies the LDP constraint. Extensive experiments are performed to validate the effectiveness and efficiency of our proposed algorithms using real datasets.

Research Full paper - Computer Science (CS) education is increasingly becoming popular across the globe. With the increased popularity of CS and Information and Communications Technology (ICT) courses, the ability to measure the similarity/distance between units can aid the decision-making process of education providers during learning path recommendations, evaluating unit exemptions/advanced standings, etc. The ability to quantitatively measure the similarity of two units will assist decision-making processes which are otherwise likely to be tedious and time-consuming. Therefore, in this work, we explore how to quantitatively measure the similarity/distance between two CS/ICT units, in an Australian setting. In this study, we utilize data from multiple CS/ICT courses offered by two Australian education providers. The data is generated as a result of the Australian Computer Society's (ACS) course accreditation process during which each unit of a course is mapped into multiple knowledge areas. We measured the similarity between units in terms of their coverage of the knowledge areas using multiple distance measures. Our work is novel in examining a quantitative measure of unit similarity, by introducing a unified approach across education providers that follow the ACS course accreditation process.

In this demonstration, we present a web based system for the novel contact tracing query (CTQ) that finds users who have come into direct contact with the query user or indirect contact via the already contacted users from a large spatio-temporal database. The CTQ is of paramount importance in the era of new COVID-19 pandemic world for identifying people who came into close spatial and temporal proximity with persons carrying an infectious disease. We demonstrate a multi-level index named QzR-tree, that considers the space coverage and the co-visiting patterns of the trajectories to group users who are likely to meet. More specifically, we use a quadtree to partition user movement traces along with a linear ordering and use the space-time mapping to group users with an R-tree. We develop a web-based demo system to show the effectiveness of the QzR-tree for the CTQ. The web-based system essentially uses a PostgreSQL database to store user trajectories, and indexes these trajectories using the QzR-tree, and finally uses a web interface to take user query and display the results in a map.

Crowd counting problem that counts the number of people in an image has been extensively studied in recent years. In this paper, we introduce a new variant of crowd counting problem, namely categorized crowd counting, that counts the number of people sitting and standing in a given image. Categorized crowd counting has many real-world applications such as crowd monitoring, customer service, and resource management. The major challenges in categorized crowd counting come from high occlusion, perspective distortion and the seemingly identical upper body posture of sitting and standing persons. Existing density map based approaches perform well to approximate a large crowd, but lose important local information necessary for categorization. On the other hand, traditional detection-based approaches perform poorly in occluded environments, especially when the crowd size gets bigger. Hence, to solve the categorized crowd counting problem, we develop a novel attention-based deep learning framework that addresses the above limitations. In particular, our approach works in three phases: i) We first generate basic detection based sitting and standing density maps to capture the local information; ii) Then, we generate a crowd counting based density map as global counting feature; iii) Finally, we have a cross-branch segregating refinement phase that splits the crowd density map into final sitting and standing density maps using attention mechanism. Extensive experiments show the efficacy of our approach in solving the categorized crowd counting problem.

In this article, we propose novel techniques to predict a user’s movie genre preference and rating behavior from her psycholinguistic attributes obtained from the social media interactions. The motivation of this work comes from various psychological studies that demonstrate that psychological attributes such as personality and values can influence one’s decision or choice in real life. In this work, we integrate user interactions in Twitter and IMDb to derive interesting relations between human psychological attributes and their movie preferences. In particular, we first predict a user’s movie genre preferences from the personality and value scores of the user derived from her tweets. Second, we also develop models to predict user movie rating behavior from her tweets in Twitter and movie genre and storyline preferences from IMDb. We further strengthen the movie rating model by incorporating the user reviews. In the above models, we investigate the role of personality and values independently and combinedly while predicting movie genre preferences and movie rating behaviors. We find that our combined models significantly improve the accuracy than that of a single model that is built by using personality or values independently. We also compare our technique with the traditional movie genre and rating prediction techniques. The experimental results show that our models are effective in recommending movies to users.

Given a set of obstacles in 2D or 3D space, a set of n candidate locations where facilities can be established, the Maximum Visibility Facility Selection (MVFS) query finds k out of the n locations, that yield the maximum visibility coverage of the data space. Though the MVFS problem has been extensively studied in visual sensor networks, computational geometry, and computer vision in the form of optimal camera placement problem, existing solutions are designed for discretized space and only work for MVFS instances having a few hundred facilities. In this paper, we revisit the MVFS problem to support new spatial database applications like "where to place security cameras to ensure better surveillance of a building complex?" or "where to place billboards in the city to maximize visibility from the surrounding space?". We introduce the concept of equivisibility triangulation to devise the first approach to accurately determine the visibility coverage of continuous data space from a subset of the facility locations, which avoids the limitations of discretizing the data space. Then, we propose an efficient graph-theoretic approach that exploits the idea of vertex separators for efficient exact in-memory solution of the MVFS problem. Finally, we propose the first external-memory based approximation algorithm (with a guaranteed approximation ratio of 1 - 1/e) that is scalable for a large number of obstacles and facility locations. We conduct extensive experimental study to show the effectiveness and efficiency of our proposed algorithms.

Deep learning based assistive technologies for the visually impaired and blind people have gained increasing attention from various research communities in recent years. In this paper, we have developed a camera-based automatic currency recognizer for Bangladeshi bank notes that assists visually impaired people in Bangladesh. We have exploited the deep learning architecture MobileNet for classification of bank notes. We have evaluated the performance of our model using a novel dataset consisting of nearly 8000 images of Bangladeshi bank notes. To verify the effectiveness and efficacy of the proposed solution, we have developed a mobile Android application, and evaluated and validated the application with the users from a blind community. The validation shows that our proposed system is robust and highly effective with heterogeneous environment.

Basic human values represent what we think are important to our lives that include security, independence, success, kindness, and pleasure. Each of us holds different values with different degrees of importance. Existing studies show that values of a person can be identified from their social network usage. However, the value priority of a person may change over time due to different factors such as time, event, influence, social structure, and technology. In this research, we are the first to investigate whether the change of value priorities can be identified from social network usage. We propose a weighted hybrid time‐series‐based model to capture the change of values of a social network user. We conducted an experimental study with 726 Facebook users and showed that our model accurately captures the value priority changes from the social network usage and achieves significantly higher accuracy than our baseline hidden Markov model‐based technique. We also validated the change of a user's value priorities in real life using a questionnaire‐based technique.

In this work, we propose a novel technique to determine the aesthetic score of a location from social metadata of Flickr photos. In particular, we built machine learning classifiers to predict the class of a location where each class corresponds to a set of locations having equal aesthetic rating. These models are trained on two empirically build datasets containing locations in two different cities (Rome and Paris) where aesthetic ratings of locations were gathered from TripAdvisor.com. In this work we exploit the idea that in a location with higher aesthetic rating, it is more likely for an user to capture a photo and other users are more likely to interact with that photo. Our models achieved as high as 79.48% accuracy (78.60% precision and 79.27% recall) on Rome dataset and 73.78% accuracy(75.62% precision and 78.07% recall) on Paris dataset. The proposed technique can facilitate urban planning, tour planning and recommending aesthetically pleasing paths.

With the widespread use of GPS-enabled mobile devices, an unprecedented amount of trajectory data has become available from various sources such as Bikely, GPS-wayPoints, and Uber. The rise of smart transportation services and recent break-throughs in autonomous vehicles increase our reliance on trajectory data in a wide variety of applications. Supporting these services in emerging platforms requires more efficient query processing in trajectory databases. In this paper, we propose two new coverage queries for trajectory databases: (i) k Best Facility Trajectory Search ( k BFT); and (ii) k Best Coverage Facility Trajectory Search ( k BCovFT). We propose a novel index structure, the Trajectory Quadtree (TQ-tree) that utilizes a quadtree to hierarchically organize trajectories into different nodes, and then applies a z-ordering to further organize the trajectories by spatial locality inside each node. This structure is highly effective in pruning the trajectory search space, which is of independent interest. By exploiting the TQ-tree, we develop a divide-and-conquer approach to efficiently process a k BFT query. To solve the k BCovFT, which is a non-submodular NP-hard problem, we propose a greedy approximation. We evaluate our algorithms through an extensive experimental study on several real datasets, and demonstrate that our algorithms outperform baselines by two to three orders of magnitude.

A socio spatial group query finds a group of users who possess strong social connections with each other and have the minimum aggregate spatial distance to a meeting point. Existing studies limit to either finding the best group of a fixed size for a single meeting location, or a single group of a fixed size w.r.t. multiple locations. However, it is highly desirable to consider multiple locations in a real-life scenario in order to organize impromptu activities of groups of various sizes. In this paper, we propose Top k Flexible Socio Spatial Group Query (Top k-FSSGQ) to find the top k groups w.r.t. multiple POIs where each group follows the minimum social connectivity constraints. We devise a ranking function to measure the group score by combining social closeness, spatial distance, and group size, which provides the flexibility of choosing groups of different sizes under different constraints. To effectively process the Top k-FSSGQ, we first develop an Exact approach that ensures early termination of the search based on the derived upper bounds. We prove that the problem is NP-hard, hence we first present a heuristic based approximation algorithm to effectively select members in intermediate solution groups based on the social connectivity of the users. Later we design a Fast Approximate approach based on the relaxed social and spatial bounds, and connectivity constraint heuristic. Experimental studies have verified the effectiveness and efficiency of our proposed approaches on real datasets.

Basic Human Values represent a set of values such as security, independence, success, kindness, and pleasure what we think are important to our lives. Each of us holds different values with different degrees of importance. Existing studies show that values of a person can be identified from her social network usage. However, the value priority of a person may change over time due to different factors such as time, event, influence, social structure, and technology. In this research, we are the first to investigate whether the change of value priorities can be identified from social network usage. We have proposed a HMM based technique to capture the change of values of a person.

In this demonstration, we present VizQ, an efficient, scalable, and interactive system to process and visualize a comprehensive collection of novel visibility queries in the presence of obstacles in 3D space. Specifically, we demonstrate four types of query processing: (i) k Maximum Visibility Query (kMVQ), that finds k locations with the maximum visibility of a target object (ii) Visibility Color Map (VCM), where each point in the space is assigned a color value denoting the visibility measure of the target (iii) Continuous Maximum Visibility (CMV) that continuously finds the location that provides the best view of a moving target, and (iv) Text Visibility Color Map (TVCM), where VCM is generated considering readability of text data displayed on a target. We are the first to propose efficient algorithms to run all of the above four types of visibility queries in the context of a large number of 3D obstacle database. We exploit human visibility metrics to design our data structures and algorithms to efficiently process queries, and our approaches outperform baseline approaches in several order of magnitude both in terms of I/Os and processing time. The link of our demonstration video is https://youtu.be/rcizJtFvQfU.

We address the problem of efficient maintenance of the answer to a new type of query: Continuous Maximizing Range- Sum (Co-MaxRS) for moving objects trajectories. The traditional static/spatial MaxRS problem finds a location for placing the centroid of a given (axes-parallel) rectangle R so that the sum of the weights of the point-objects from a given set O inside the interior of R is maximized. However, moving objects continuously change their locations over time, so the MaxRS solution for a particular time instant need not be a solution at another time instant. In this paper, we devise the conditions under which a particular MaxRS solution may cease to be valid and a new optimal location for the query-rectangle R is needed. More specifically, we solve the problem of maintaining the trajectory of the centroid of R. In addition, we propose efficient pruning strategies (and corresponding data structures) to speed-up the process of maintaining the accuracy of the Co-MaxRS solution. We prove the correctness of our approach and present experimental evaluations over both real and synthetic datasets, demonstrating the benefits of the proposed methods.

The rise of innovative transportation services and the recent breakthrough in the development of autonomous vehicles have stimulated the research on collective travel planning problems such as ride-sharing, carpooling, and on-demand vehicle routing in recent years. In this paper, we introduce several optimization problems to recommend a suitable route and stops of a vehicle, in a road network, for a group of users intending to travel collectively. The goal of each problem is to minimize the aggregate cost of the individual travelers' paths and the shared route under various constraints. First, we introduce the optimal end-stops (OES) query that finds a pair of pick-up-and-drop-off locations such that the sum of the distance between these locations and the total distance traveled by the travelers from their start locations to the pick-up location and from the drop-off location to their end locations is minimized. We propose a polynomial-time fast algorithm for the OES query by utilizing the path-coherence property of road networks. Second, we formulate the optimal route and intermediate stops (ORIS) query to find a set of intermediate stops for the vehicle such that the sum of the total distance traveled by the vehicle and the total distance traveled by the travelers from their start locations to one of the stops and to their end locations from one of the stops is minimized. We propose a novel near-optimal polynomial-time-and-space heuristic algorithm for the ORIS query that performs reasonably well in practice. We also analyze several variants of this problem. Finally, we perform extensive experiments to demonstrate the efficiency and efficacy of our algorithms.

Reducing the road traffic congestion has become an important challenge in recent years; the researchers have focused on identifying causes and remedies for the traffic congestion. However, the impact of the locations and activity time of points of interests (POIs) on the road traffic has not yet been explored. Moreover, in developing countries, POIs are not established in planned manner and cause the traffic congestion. In this paper, we analyze how POI locations and activities affect the road traffic. Specifically, we identify the congestion pattern caused by POI activities in the field study and develop a model to quantify the spatio-temporal impact of POI activities on the traffic congestion. Moreover, we develop a model to predict the impact of POI activities on the road traffic congestion using fuzzy regressions for different POI categories. We perform a set of case studies and experiments to show the effectiveness of our models using real datasets of POI activities and road traffic.

Traffic jam in Dhaka city, the capital of Bangladesh and one of the most densely populated cities in the world, is one of the major problems of the commuters. The city dwellers have been experiencing intolerable traffic jam everyday. To reduce the traffic congestion, new roads have been built and new rules for vehicles in the roads have been enforced. These changes do not result in any visible improvement of the traffic condition. One major bottleneck for not being able to take appropriate measure for combating traffic jam is the inability to assess and predict the traffic condition of the road. Due to huge infrastructural cost, the city does not have facility to collect traffic data by deploying sensors. In this paper, we overcome this limitation by exploiting mobile phone call details record (CDR) data collected by mobile phone operators for billing purpose. We propose a methodology to estimate travel time between any two major locations of Dhaka city from the aggregated information of a high volume of users' mobile phone CDR data. Our experimental results on real CDR data set of 2.87 millions of users of the largest mobile operator, Grameenphone Ltd., show that we can effectively predict travel time between any two major junctions and any two minor locations of the city with an average accuracy of 87% and 76% respectively.

We address the problem of maintaining the correct answer-sets to the Conditional Maximizing Range-Sum (C-MaxRS) query in spatial data streams. Given a set of (possibly weighted) 2D point objects, the traditional MaxRS problem determines an optimal placement for an axes-parallel rectangle r so that the number -- or, the weighted sum -- of objects in its interior is maximized. In many practical settings, the objects from a particular set -- e.g., restaurants -- can be of distinct types -- e.g., fast-food, Asian, etc. The C-MaxRS problem deals with maximizing the overall sum, given class-based existential constraints, i.e., a lower bound on the count of objects of interests from particular classes. We first propose an efficient algorithm to the static C-MaxRS query, and extend the solution to handle dynamic (data streams) settings. Our experiments over datasets of up to 100,000 objects show that the proposed solutions provide significant efficiency benefits.

Street crime is a prevalent problem in developing countries like Bangladesh. Though this problem has been identified long before, no visible remedy or action can be seen to overcome or combat these street crimes in one of the most populated mega-cities, Dhaka, Bangladesh, of the world. In this paper, we propose a novel spatio-temporal street crime prediction model that exploits the historical street crime data of Dhaka city to predict the possibility of a crime in a particular region at a specific time. Our model captures both space and time proximity of past crimes while predicting a future crime. Experimental evaluation shows that our spatiotemporal prediction model can predict a future crime with 79.24% sensitivity and 68.2% specificity. As a proof of concept we develop an Android application that alerts a user about the possibility of different crimes in a place at a particular time.

Contemporary modern world has witnessed the widespread emergence of online social media and similar technologies. Peoples' behaviour over different social network platform has become an interesting topic of research. In this study, we investigate whether people express analogous identity over different platforms and analysis of different social platform usage contributes to reveal more of a person. We analyse people's usage pattern in two major online platforms, the most widely used social media platform Twitter and a major online commenting platform Disqus. We extract linguistic features and infer personality traits from both of these platforms. Our study reveals differential relationship between personality traits and Disqus and Twitter usage. We also find that social media has an influence on a person's discussion topic. People share opinion on varieties of topics and entities exclusively in Twitter and Disqus. Moreover Disqus provides stronger assessment of a person's sentiment over a topic or entity. Combination of these two profiles gives an extensive view of a user's interest and sensitivity which justify the inference that people use different social network for different purposes and single social network analysis is not enough to build a comprehensive virtual identity of a person.

A group of friends located at their working places may want to plan a trip to visit a shopping center, have dinner at a restaurant, watch a movie at a theater, and then finally return to their homes with the minimum total trip distance. For a group of spatially dispersed users a group trip planning (GTP) query returns points of interests (POIs) of different types such as a shopping center, a restaurant and a movie theater that minimize the aggregate trip distance for the group. The aggregate trip distance could be the sum or maximum of the trip distances of all users in the group, where the users travel from their source locations via the jointly visited POIs to their individual destinations. In this paper, we develop both optimal and approximation algorithms for GTP queries for both Euclidean space and road networks. Processing GTP queries in real time is a computational challenge as trips involve POIs of multiple types and computation of aggregate trip distances. We develop novel techniques to refine the POI search space for a GTP query based on geometric properties of ellipses, which in turn significantly reduces the number of aggregate trip distance computations. An extensive set of experiments on a real and synthetic datasets shows that our approach outperforms the most competitive approach on an average by three orders of magnitude in terms of processing time.

Sexual harassment of women in public places (e.g., foot-paths, buses, and shopping malls) of major cities in developing countries is a growing concern. These harassments can happen in various forms ranging from commenting, catcalling, and staring to touching and groping, to attacking and raping. Though, the most severe form of harassments such as attacking and raping get some attention from the society, NGOs and law-enforcement agencies, unfortunately, other forms of harassments that are more widespread in public places remain largely un-attended or ignored in our conservative society. However, these harassments are more common and can have various negative psychological impacts on women that include a persistent feeling of insecurity, loss of self-esteem, restricted participation in daily life activities in public places. In this paper, we propose a crowd-powered privacy-aware location based mobile application, SafeStreet, that empowers women in public places against sexual harassments. SafeStreet allows a women to privately capture and share her own experiences in the street. SafeStreet enables a women to find a safe path, i.e., the path to a destination that has less harassment hazard, at any point of time.

The recent advancement of GPS-enabled mobile technologies and the proliferation of map-based applications are attracting an increasing number of people to use location based services (LBSs). Processing a larger number of simultaneous queries efficiently have become an important research topic in recent years. In this paper, we focus on an important class of LBSs, shortest path queries (SP-queries) in road networks. Given a source and a destination in a road network, an SP-query returns the path from the source to the destination that minimizes the travel time. We particularly focus on batch processing of simultaneous SP-queries in road networks. Traditional systems that process one query at a time usually provide slow responses, causing the machine to flood with incoming queries. Existing fast solutions for SP-queries require expensive pre-processing steps and are incapable of adapting with the continuous change in traffic on the roads. We propose an efficient group based approach that provides an approximate solution with reduced cost and high accuracy. An important benefit of our approach is that it does not require expensive pre-processing. The key concept is to exploit the path-coherence property of road networks by grouping queries that share substantial common paths in their shortest paths and processing the group in a single pass. Our approach incurs an average relative error of 0.5% and is on average 6 times faster than the straightforward approach that evaluates each SP-query individually.

The proliferation of location-based social networks allows people to access location-based services as a group. We address Group Optimal Sequenced Route (GOSR) queries that enable a group to plan a trip with a minimum aggregate trip distance. The trip starts from the source locations of the group members, goes via a predefined sequence of different point of interests (POIs) such as a restaurant, shopping center and movie theater, and ends at the destination locations of the group members. The aggregate trip distance can be the total or the maximum trip distance of the group members. We introduce a novel approach to efficiently compute group optimal sequenced routes in road networks. We exploit elliptical properties to refine the POI search space and develop efficient algorithms for GOSR queries. Experiments show that our approach outperforms a naive approach significantly in terms of processing time and I/Os.

User privacy in location-based services (LBSs) has become an important research area. We introduce a new direction to protect user privacy that evaluates LBSs with crowdsourced data and computation and eliminates the role of a location-based service provider. We focus on the group nearest neighbor (GNN) query that allows a group to meet at their nearest point of interest such as a restaurant that minimizes the total or maximum distance of the group. We develop a crowdsource-based approach, called PrivateMeetUp, to evaluate GNN queries in a privacy preserving manner and implement a working prototype of PrivateMeetUp.

Many real-world problems, such as placement of surveillance cameras and pricing of hotel rooms with a view, require the ability to determine the visibility of a given target object from different locations. Advances in large-scale 3D modeling (e.g., 3D virtual cities) provide us with data that can be used to solve these problems with high accuracy. In this paper, we investigate the problem of finding the location which provides the best view of a target object with visual obstacles in 2D or 3D space, for example, finding the location that provides the best view of fireworks in a city with tall buildings. To solve this problem, we first define the quality measure of a view (i.e., visibility measure) as the visible angular size of the target object. Then, we propose a new query type called the k-Maximum Visibility (kMV) query, which finds k locations from a set of locations that maximize the visibility of the target object. Our objective in this paper is to design a query solution which is capable of handling large-scale city models. This objective precludes the use of approaches that rely on constructing a visibility graph of the entire data space. As a result, we propose three approaches that incrementally consider relevant obstacles in order to determine the visibility of a target object from a given set of locations. These approaches differ in the order of obstacle retrieval, namely: query centric distance based, query centric visible region based, and target centric distance based approaches. We have conducted an extensive experimental study on real 2D and 3D datasets to demonstrate the efficiency and effectiveness of our solutions.

Continuous queries in road networks have gained significant research interests due to advances in GIS and mobile computing. Consider the following scenario: "A driver uses a networked GPS navigator to monitor five nearest gas stations in a road network." The main challenge of processing such a moving query is how to efficiently monitor network distances of the k nearest and possible resultant objects. To enable result monitoring in real-time, researchers have devised techniques which utilize precomputed distances and results, e.g., the network Voronoi diagram (NVD). However, the main drawback of preprocessing is that it requires access to all data objects and network nodes, which means that it is not suitable for large datasets in many real life situations. The best existing method to monitor kNN results without precomputation relies on executions of snapshot queries at network nodes encountered by the query point. This method results in repetitive distance evaluation over the same or similar sets of nodes. In this paper, we propose a method called the local network Voronoi diagram (LNVD) to compute query answers for a small area around the query point. As a result, our method requires neither precomputation nor distance evaluation at every intersection. According to our extensive analysis and experimental results, our method significantly outperforms the best existing method in terms of data access and computation costs.

With recent advances in mobile computing technologies, mobile devices can now render 3D objects realistically. Many users of these devices such as tourists, mixed-reality gamers, and rescue officers, need real-time retrieval of 3D objects over a wireless network. Due to bandwidth and latency restrictions in mobile settings, efficient continuous retrieval of 3D objects remains a challenge. In this paper, we describe a motion-aware approach to this problem. We first introduce multi-resolution storage and retrieval methods for 3D data, which restrict access to only the necessary content based on the client's motion pattern. We then propose a motion-aware buffer management technique as well as an efficient index using multi-resolution representations of objects. Our experiments demonstrate the effectiveness of our solution to continuous retrieval of complex spatial data in mobile settings.