University of Twente Student Theses


Personalised Coaching of Patients with Type 2 Diabetes Based on Their Individual Behavioural Characteristics

Nyhoff, Tom (2023) Personalised Coaching of Patients with Type 2 Diabetes Based on Their Individual Behavioural Characteristics.

[img] PDF
Abstract:This research examined how statistical methods like cluster analysis can be optimally used to support the development of eHealth interventions by allowing better personalisation. More specifically, this can be done by identifying patients with similar characteristics and would therefore profit from specific coaching strategies. The three unsupervised learning methods, K-Means clustering, Self-Organising Map algorithms and Swarm Intelligence Based clustering, were chosen and compared using a dataset of Type 2 Diabetes patients from the Diabetes and Lifestyle Cohort Twente. These three methods were applied to a dataset measuring the patients’ physical activity using their step count at specific times as a measure, a dataset measuring the nutritional intake based on food diaries, and a dataset combining those two. The results of these methodologies were then evaluated based on interpretability and cluster quality. Interpretability was measured by analysing the number of variables showing significant differences between groups and how relevant these variables are for Type 2 Diabetes based on the literature. Mean silhouette scores were used as a measure of cluster quality. Overall, the results suggest that K-means clustering is a valid choice overall, as its results for the silhouette scores suggest the best cluster quality of the three methods.
Item Type:Essay (Bachelor)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:77 psychology
Programme:Psychology BSc (56604)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page