@misc{a55aaef106fc4fb3bd908fcca56e4489,
title = "Spatial Data Science: Applications and Implementations in Learning Human Mobility Patterns",
abstract = "In this modern era, many devices and sensors collect data with spatial and temporalcharacteristics. This type of data is referred to as spatio-temporal data and can be usedto derive valuable knowledge about the mobility patterns of people. As a result, differentorganizations are increasingly interested in using this knowledge to make a social impactand improve people{\textquoteright}s lives.The process of extracting knowledge and discovering patterns in data fall within therealm of data mining. This thesis presents two data mining applications, focusing ontrajectory data, a specific type of spatio-temporal data related to the movement tracesof moving objects, especially humans and vehicles. The two applications presented aimto derive people{\textquoteright}s mobility patterns and gain insights about these, making it possible tomake informed decisions.The first application focuses on detecting wandering behavior in people with dementia by identifying the problem as a trajectory anomaly detection problem. A systemarchitecture with an integrated anomaly detection approach is developed and implemented to analyze the real-time movement behavior of a person with dementia to detect their anomalous trajectory. The system can also initiate and manage a rescue mission involving relatives and volunteers if the detection result is positive. The applicationaims to alleviate relatives and help persons with dementia in case they are in need.The second application focuses on different approaches to estimating travel demandin urban traffic networks based on observations, at certain locations, of vehicles that traverse the roads of a traffic network. In particular, the focus is on a bi-level optimizationproblem formulation: The dynamic origin-destination problem, which is especiallycomputationally expensive to solve. For this reason, a new (surrogate model-based) approach is proposed and compared against existing classical approaches to determine thebest and highlight different advantages and disadvantages. The application aims to develop and determine the best approach for travel demand estimation, which is essentialwhen improving transportation facilities and infrastructure.The overall contributions are highlighted for each application presented, and variousopen problems and future work are discussed.",
keywords = "geospatial, spatial, temporal, spatio-temporal, data mining, trajectory, optimization, transportation, estimation, dementia, anomaly detection, data stream, online, offline, real-time, system architecture, software development",
author = "Andersen, {Nicklas Sindlev}",
year = "2022",
month = dec,
day = "19",
doi = "10.21996/kt5a-3r71",
language = "English",
publisher = "Syddansk Universitet. Det Naturvidenskabelige Fakultet",
address = "Denmark",
school = "SDU",
}