Machine Learning Regression Tree Approach for Age Prediction from Eruption Status of Permanent Teeth in Sri Lankan Children

Abstract

Prediction of age is a predominant facet in forensic and clinical fields. Forensic odontology is used to predict an age by using permanent teeth resistant to high temperatures and any mass disaster than other parts of the body. Although many studies have been carried out in other countries, this is the first study of age prediction using the eruption status of permanent teeth for Sri Lankans. Age-related memory loss, memory loss associated with dementia, and the absence of official documents to verify their age are the main reasons people have no knowledge about their age. Therefore, age prediction is used in various situations, such as identification, admission purposes, employment, criminal issues and judicial punishments. The main objective of this study is to predict the age of a child using the eruption status of permanent teeth. This cross-sectional study was conducted on 3321 individuals (1681 males and 1640 females) from 7 provinces and 20 schools in Sri Lanka. Regression tree algorithms in Machine learning were used for age prediction. Classification and regression trees (CART), gradient boosting (GB) classifier and extreme gradient boost (XGBoost) classifier were used to make predictions for the age of a  child. Results were validated using cross-validation techniques, and root mean squared error (RMSE) and R-squared values were used as accuracy measures to select the best model. The best model for age prediction was the XGBoost model, which gives the highest accuracy (88%). This is the first study of age prediction using eruption status of permanent teeth for Sri Lankan children. The study results provide an XGBoost machine learning classifier as the most suitable method for age prediction with higher precision.