OpenTS-Applications
AIOps
Auto-scaling of cloud resources based on probabilistic future workload forecasting. The related research has been successfully implemented in Alibaba Cloud.
MagicScaler: Uncertainty-Aware, Predictive Autoscaling
International Conference on Very Large Databases (PVLDB), 2024.
Intelligent Transportation
Path planning and recommendation with dynamic, uncertain road network modeling with uncertain traffic time series. The related research has been successfully implemented at FlexDanmark.
Efficient Stochastic Routing in Path-Centric Uncertain Road Networks
International Conference on Very Large Databases (PVLDB), 2024.
Stochastic Routing with Arrival Windows
ACM Transactions on Spatial Algorithms and Systems (TSAS), 2023.
Context-aware Path Ranking in Road Networks
IEEE Transactions on Knowledge and Data Engineering(TKDE), 2020.
Anytime Stochastic Routing with Hybrid Learning
International Conference on Very Large Databases (PVLDB), 2020.
Context-aware, preference-based vehicle routing
The International Journal on Very Large Data Bases (VLDBJ), 2020.
Virtual Power Plants (VPP)
Leveraging AI4TS to drive Virtual Power Plants by optimizing the aggregation, dispatch, and market trading of massive distributed energy resources.
AI4DB
Root cause analysis of slow queries in cloud databases based on multi-modal observation data (including metrics time-series, SQL statements, logs, and execution plans). The related research has been successfully implemented in Alibaba Cloud.
RCRank: Multimodal Ranking of Root Causes of Slow Queries in Cloud Database Systems
International Conference on Very Large Databases (PVLDB), 2025.
FinTech
By leveraging AI4TS, FinTech transforms vast amounts of financial data into actionable insights, driving innovations in precise risk management, personalized services, and automated decision-making to enhance efficiency, security, and customer experience.
Case 1: In collaboration with Zhonghui Information Technology (Shanghai) Co., Ltd., we proposed a price prediction model for foreign exchange spot products based on hypergraph networks and multi-scale learning. When applied to the foreign exchange transaction monitoring environment, it improved the accuracy of abnormal transaction price judgment and traceability.
Introduction to Correlation
Smart Emergency Response
AI4TS empowers smart emergency response by fusing multi-modal data—including the Internet of Things, satellite images, InSAR data, and social media—for predictive analytics, dynamic situational awareness, and optimized resource allocation, ultimately enabling faster, more precise, and more effective life-saving interventions.
Case 1: In collaboration with JD.com and Suqian Housing and Urban-Rural Development Bureau, we developed the physical knowledge-guided deep learning model Waterlogformer, which was implemented on the “Suqian City Lifeline Safety Supervision Platform” to provide early warning services for urban waterlogging in Suqian.
Groundtruth
Waterlogformer
Case 2: In collaboration with the Department of Natural Resources of Guizhou Province, we developed the Multi-modal Adaptive Association Learning (MAAL) framework, which was implemented on the Guizhou Geological Hazard Prevention and Control Command Platform. This framework covers 889 geological hazard risk areas in the 176,167 square kilometers of mountainous areas in Guizhou Province, and conducts deformation trend prediction.
Comparison between MAAL model deformation prediction results and measured data
Case 3: In collaboration with JD.com and the Pudong New Area Emergency Management Bureau, we have developed an intelligent urban emergency safety body that provides sales staff with full-process risk assessment and disposal suggestions by accessing real-time data such as weather and road conditions, GIS information around emergency events, and knowledge related to emergency plans. It covers the entire life cycle of emergency event disposal, from risk identification and disposal suggestions to review and optimization.
Urban emergency safety intelligence