Keywords
Orthodontic, machine learning, transparency, explainability, SHapley Additive exPlanations
Abstract
The study aims to answer the research question: how can machine learning models be optimized to enhance the accuracy and transparency of Invisalign treatment outcome predictions? The research methodology deployed in this study focuses on applying the Design Science Research and Cross Industry Standard Process for Data Mining (CRISP-DM) to construct a predictive model for Invisalign treatment outcomes. The data set is from five distinct private dental clinics across Thailand, consisting of 657 de-identified orthodontic treatment records. Machine learning models, Decision tree, random forest, Neural Net and XGBoost were selected for the model development to ensure transparency and interpretability. SHapley Additive exPlanations (SHAP) analysis of ANN and XG Boost demonstrated the feature importance in Invisalign treatment decision interpretability.
Recommended Citation
Win, Khin Than; Trakulmututa, Sanisa; and Fahmideh, Mahdi, "Orthodontic Predictions with Machine Learning" (2024). Digit 2024 Proceedings. 20.
https://aisel.aisnet.org/digit2024/20