AI-Driven Automation of Supplement Recommendations

Sammanfattning

This thesis focuses on the utilization of machine learning models to automate supplement recommendations based on wellness screening data collected at a healthcare clinic. The problem posed concerns how AI could use health markers to reproduce the recommendation process normally carried out by a healthcare professional. The project was driven by the need for more effective, easier to use, and widely available preventive health care services. To achieve this goal, three models of machine learning were applied: Random Forest, Support Vector Machine, and XGBoost. The models were used to create 28 binary recommendation predictions based on 38 health marker inputs. Performance evaluation was done using F1-score and exact match accuracy. Focusing on generalization rather than relying on vast amounts of data and complex datasets, XGBoost stood out among all tested models, accomplishing an exact match accuracy of 69%. There is no doubt that the models do not achieve the level of precision typically exhibited by healthcare professionals. Nevertheless, the research shows that the application of Artificial Intelligence within healthcare has a promising future.