【工程管理与技术经济系邀请】

题目：Scalable data fusion for generating disaggregate activity schedules

主讲人：Prateek Bansal 助理教授，新加坡国立大学

时间：2026年6月9日15:30

地点：经管楼905室

摘要：Understanding individual-level mobility is critical for advancing transport and urban planning. Yet, no single dataset offers both behavioral richness and spatiotemporal coverage: household travel surveys capture detailed demographic and behavioral information but lack spatial and temporal completeness, whereas passively collected data (e.g., mobile or sensor traces) offer extensive coverage but omit demographic and contextual details. We propose a two-stage data fusion framework that integrates these complementary sources. In the first stage, a cluster-based encoder–decoder model aligns sociodemographic, behavioral, and spatiotemporal features across datasets. In the second stage, a context-aware Markov process generates realistic daily activity schedules consistent with both micro-level behavior and population-level mobility patterns. The framework produces behaviorally realistic and spatially diverse activity patterns, correcting survey biases (e.g., commuting overrepresentation) and expanding spatiotemporal coverage by over 100-fold. Applied to Singapore and Seoul, it provides high-fidelity demand-side behavioral data to support adaptive transport and urban planning.

个人简介：Prateek Bansal 2019年毕业于康奈尔大学，现任新加坡国立大学土木与环境工程系助理教授，现领导新加坡国立大学行为认知科学实验室（Behavioural Cognitive Science Lab），并担任未来城市实验室（Future Cities Laboratory）Adaptive Mobility模块的共同首席研究员。主要研究领域为贝叶斯机器学习、计量经济学与计算心理学交叉领域，专注于个体行为研究与城市尺度交通、能源系统设计。担任Transportation Research Part A、B副主编，Transportation Research Part C编委，围绕AI+统计机器学习、交通科学、能源行为与能源系统等在Nature Communications， Transportation Research B等高水平期刊发表论文80余篇。