Periodontal disease affects over 1.1 billion individuals globally, yet conventional risk assessment methods inadequately capture complex relationships between oral hygiene behaviors and disease outcomes. To implement and validate machine learning algorithms for predicting periodontal disease risk through comprehensive analysis of oral hygiene patterns, clinical parameters, and demographic determinants. The study analyzed 409 patients from the University of Baghdad College of Dentistry (July 2024-May 2025). Data included clinical assessments (plaque index, bleeding on probing, clinical attachment loss, probing pocket depth), demographics, anthropometrics, and behavioral indicators. Four algorithms were evaluated: Random Forest, Gradient Boosting, Logistic Regression, and Support Vector Machine using stratified cross-validation. The cohort comprised 150 periodontitis (36.7%), 147 gingivitis (35.9%), and 112 healthy patients (27.4%). Brushing frequency showed a 4.7-fold difference between healthy (2.27±0.63 times/day) and periodontitis patients (0.48±0.55 times/day). Strong correlations emerged between brushing frequency and plaque index (r=-0.845), bleeding on probing (r=-0.800), and clinical attachment loss (r=-0.325). Random Forest achieved 100% accuracy with 99.6%±0.7% cross-validation reliability. Machine learning algorithms demonstrate exceptional capability for periodontal disease prediction and risk stratification, establishing foundations for precision-based clinical decision support and personalized intervention strategies.
