Thesis
A preliminary investigation of a personalized risk alert system for weight control lapses
Master of Science (M.S.), Drexel University
Mar 2016
DOI:
https://doi.org/10.17918/etd-6683
Abstract
Failures to adhere to dietary recommendations in a weight loss program have been attributed to the notion that short-term violations of behavioral "rules" (e.g., dietary lapses) can often precipitate a return to prior behavior patterns. Existing studies suggest that lapses are caused by a select group of internal and external cues, indicating that they may be predictable. Just-in-time adaptive interventions (JITAI) can utilize mathematical models to learn the way in which triggers predict lapse behavior, and then communicate this knowledge to the individual in the form of momentary alerts to risk. The primary aim of the current study was to develop an initial machine learning model (accuracy > 70%, sensitivity > 70%, and specificity > 50%) capable of predicting lapse behavior in a sample of overweight and obese participants following a standardized weight control diet. Users (n = 12, MBMI = 33.6) were prompted to report on lapses and relevant triggers six quasi-random times per day via personal smartphone for six weeks. Participants reported an average of 3.47 lapses per week (SD = 2.41) and completed an average of 94.64% of prompts. This data was used to build a classification model (using C4.5 WEKA decision trees) to predict dietary lapses. The final model accuracy (.72) met our standards for success while maintaining good sensitivity (.70) and specificity (.72). Significance testing revealed this model predicted lapses better than random chance (p < .01). Further analyses related to data collection procedures, variable selection, and individual-level models are presented below. Results from this project indicate that machine learning holds promise for the real-time prediction of dietary lapse, and the algorithm from this project will be utilized in a JITAI application framework to predict and prevent lapses.
Metrics
68 File views/ downloads
63 Record Views
Details
- Title
- A preliminary investigation of a personalized risk alert system for weight control lapses
- Creators
- Stephanie Paige Goldstein - DU
- Contributors
- Evan M. Forman (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- ix, 77 pages
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- Psychological and Brain Sciences (Psychology); College of Arts and Sciences; Drexel University
- Other Identifier
- 6683; 991014632927104721