University of Twente Student Theses
Developing a Recommendation Algorithm for Patients Using the Healthentia Platform
Bujorianu, Alex (2022) Developing a Recommendation Algorithm for Patients Using the Healthentia Platform.
PDF
1MB |
Abstract: | The purpose of this work is to develop a high-level recommendation algorithm that provides medical advice to users of the Healthentia Virtual Coaching platform. Training data is generated using a proprietary simulation part-developed by the author, and manually labelled. A variety of machine learning algorithms are trained and their performance evaluated on the training set using cross-validation. In addition, a “knowledge algorithm” is created by formalising domain knowledge rules. The results indicate that the knowledge algorithm performs the best. In terms of its contribution to the literature, this work emphasises the fact that rules-based knowledge algorithms can perform better than learning-based approaches, provided that the problem is well-understood by domain experts and formalisable. Given the disadvantages of supervised learning-based approaches – sensitivity to unbalanced data, labelling requirements, overfitting, training/optimisation time, and sensitivity to missing data – knowledge-based algorithms provide a compelling alternative. This is particularly true of the medical field, where labelling often needs to be done by experts, and data is frequently incomplete. |
Item Type: | Essay (Master) |
Clients: | Innovation Sprint, Brussels, Belgium |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 44 medicine, 54 computer science, 77 psychology |
Programme: | Business Information Technology MSc (60025) |
Link to this item: | https://purl.utwente.nl/essays/93012 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page