Predictive analytics over mobility data are of great importance since they can assist an analyst to predict events, such as collisions, encounters, traffic jams, etc. A typical example of such analytics is future location prediction, where the goal is to predict the future location of a moving object,given a look-ahead time. What is even more challenging is being able to accurately predict collective behavioural patterns of movement, such as co-movement patterns. In this paper, we provide an accurate solution to the problem of Online Prediction of Co-movement Patterns. In more detail, we split the original problem into two sub-problems, namely Future Location Prediction and Evolving Cluster Detection. Furthermore, in order to be able to calculate the accuracy of our solution, we propose a co-movement pattern similarity measure, which facilitates us to match the predicted clusters with the actual ones. Finally, the accuracy of our solution is demonstrated experimentally over a real dataset from the maritime domain.
Learning-based Online Optimization for Autonomous Mobility-on-Demand Fleet Control
TUM School of Management(慕尼黑工业大学管理学院);Technical University of Munich(慕尼黑工业大学);CERMICS(法国桥路学校计算数学与科学工程中心);École des Ponts(法国桥路学校);Munich Data Science Institute(慕尼黑数据科学研究所);CIRRELT(蒙特利尔物流、运输与计算机工程研究中心);École Polytechnique de Montréal(蒙特利尔理工学院);Pontifical Catholic University of Rio de Janeiro(里约热内卢天主教大学)
BERT4Traj: Transformer Based Trajectory Reconstruction for Sparse Mobility Data
University of Georgia(佐治亚大学);Department of Geography University of Georgia(佐治亚大学地理系);College of Public Health University of Georgia(佐治亚大学公共卫生学院);University of Texas at Austin(德克萨斯大学奥斯汀分校)
MoE-TransMov: A Transformer-based Model for Next POI Prediction in Familiar & Unfamiliar Movements
Lyles School of Civil and Construction Engineering, Purdue University, West Lafayette, IN, USA(普渡大学土木与建设工程学院);LY Corporation, Tokyo, Japan(东京LY公司);Tandon School of Engineering, New York University, New York, NY, USA(纽约大学塔能工程学院)
Online Optimization with Predictions and Non-convex Losses
IIIS, Tsinghua University(清华大学交叉信息研究院);California Institute of Technology(加州理工学院)
我们研究在线优化的设置,在线学习者寻求优化每轮命中成本,这可能是非凸的,而在轮之间改变动作时会产生移动成本。我们要求:\textit{under what general conditions is it possible for an online learner to leverage predictions of future cost functions in order to achieve near-optimal costs?}先前的工作提供了接近最优的在线算法的特定组合的假设击中和切换成本,但没有一般的结果是已知的。在这项工作中,我们给出了两个一般的充分条件,指定击球和运动成本之间的关系,保证了一个新的算法,同步固定地平线控制(SFHC),提供了一个$1+O(1/w)$竞争比,其中$w$是学习者可用的预测数。我们的条件不要求成本函数是凸的,我们也得到了非凸的打击和运动成本的竞争比结果。我们的研究结果提供了第一个常数,无量纲的竞争比在线非凸优化运动成本。此外,我们给出了一个自然实例的例子,凸体追逐(CBC),其中不满足充分条件,我们可以证明没有在线算法可以具有收敛到1的竞争比。
What are People Talking about in #BlackLivesMatter and #StopAsianHate? Exploring and Categorizing Twitter Topics Emerging in Online Social Movements through the Latent Dirichlet Allocation Model
Duke Kunshan University(杜克-昆山大学);Data Science Research Center(数据科学研究中心);Social Science Division Duke Kunshan University(社会科学系杜克-昆山大学)