Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/159246
Title: Data-driven prediction and impact analysis for smart city applications
Authors: Mo, Xiaoyun
Keywords: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
Issue Date: 2022
Publisher: Nanyang Technological University
Source: Mo, X. (2022). Data-driven prediction and impact analysis for smart city applications. Doctoral thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/159246
Project: MOE Tier 1 RG18/20 
NRF SDSC-2019-001 
Abstract: Predictive analysis of discrete events in continuous time, such as incidents in public systems like the rail transit systems, and dwellers' activities of taxi ridings, clinical visits, etc., are critical to improving public services and life quality in smart cities. In this thesis, we develop three projects aiming to answer the following questions, namely, how to conduct impact analysis on abnormal events (e.g., a transit disruption), how to predict the occurrence of an abnormal event, and how to predict the occurrence of a normal event (e.g., a clinical visit). Firstly, we predict the impact of a service disruption in an urban rail transit system. We define two impact metrics and derive the predictor of each metric based on the inferred alternative route choices of commuters under disruptions. Secondly, we develop a stochastic model to predict when and where a service delay or disruption may occur in rail systems. We leverage only basic public information of events and propose a non-trivial method based on multivariate Hawkes process. Finally, we propose a general stochastic model to predict the occurrence of normal events, which is a neural temporal point process formulated by a novel mixture model of monotonic neural networks. We conduct extensive experiments on real-world datasets in these projects, and the results demonstrate the superiority of our methods to existing approaches.
URI: https://hdl.handle.net/10356/159246
Schools: School of Computer Science and Engineering 
Rights: This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).
Fulltext Permission: open
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Theses

Files in This Item:
File Description SizeFormat 
amended thesis.pdf2.62 MBAdobe PDFThumbnail
View/Open

Page view(s)

231
Updated on Jun 23, 2024

Download(s)

2
Updated on Jun 23, 2024

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.