Dissertation
POPULATION-LEVEL BEHAVIOR ANALYSIS BASED ON SMART ENVIRONMENT SENSOR DATA
Doctor of Philosophy (PhD), Washington State University
01/2020
Handle:
https://hdl.handle.net/2376/107690
Abstract
Sensor data that are collected as people perform ordinary routines in their homes provide insights about human behavior. Comparing sensor data-derived norms among population subgroups offers the potential to transform how important services are delivered in millions of homes. With massive amounts of available sensor data, we have entered an era in which a much greater understanding of complex behavior can be gained through the development of innovative algorithms to analyze such sensor data.
This dissertation focuses on population-level behavior analysis on smart environment sensor data. We leverage decades of behavioral sensor data from over 100 smart homes to identify routine behavior patterns and assess behavior changes with a view to applying our findings to personalized healthcare.
We design three methods to formally model resident behavior based on smart home sensor data. We first introduce two stochastic methods with mechanistic descriptions of behavior patterns to model resident behavior at a population level and compare behavior differences among population groups. We also design a data-driven approach, inverse reinforcement learning, to model and quantify residents’ behavior as well as to distinguish cognitively impaired groups from healthy populations. The findings will offer the potential to automate diagnoses and design customized behavioral interventions as well as inform strategies for improving people’s health-promoting daily habits.
Metrics
42 File views/ downloads
33 Record Views
Details
- Title
- POPULATION-LEVEL BEHAVIOR ANALYSIS BASED ON SMART ENVIRONMENT SENSOR DATA
- Creators
- BEIYU LIN
- Contributors
- Diane Cook (Advisor)Maureen Schmitter-Edgecombe (Committee Member)Sandip Roy (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 139
- Identifiers
- 99900581808501842
- Language
- English
- Resource Type
- Dissertation