Abstract: Objective: To establish a linear regression model for shade matching based on machine learning algorithms. Methods: One hundred and twenty-three undergraduates were recruited as the volunteers of the present study(mean age: 22.48±1.77 years old, male: female=64:59). The chromatic valuesof volunteer’s maxillary teeth (15-25) were measured with a spectrophotometer. The matrix of chromatic value (X) and regression coefficient (θ) were determined based on the data from the volunteers. Efforts have then been put into the optimizing evaluation function through collaborative filtering. Finally, a multiple linear regression model was established with the purpose to predict the chromatic value of missing tooth. Twenty-four volunteers at similar ages were sampled to evaluate the prediction accuracy of the multiple linear regression model. Thepredicted chromatic value of their teeth were calculated and compared to the measured values based on the spectrophotometer. Results: The prediction accuracy of the model on L^*value was more than 80% among 94.17% of 240 teeth, while the prediction accuracy ona^*value and b^*value were more than 80% among the entire sampling teeth. Conclusion: The multiple linear regression model might be a suitable tool to predict the color of tooth in permanent dentition.

Key words: Shade matching, Prediction, Chromatic value, Linear regression model