Abstract Background/purpose There are still cases of dental implant failure in Taiwan. In addition, this treatment involves deeper embedding, which may easily cause medical disputes. The aim of this study was to analyze implant data to generate classification rules which can be used as a predictive method prior to implant surgery. Materials and methods In total, 1161 implants from 513 patients were included in this study. Data on 23 items were collected and treated as impact factors on dental implants. In addition, information on the individual health of patients related to the 23 impact factors was collected. The 1161 implants were then analyzed using the C5.0 method to establish a prediction model. Three performance indicators of accuracy, sensitivity, and specificity were also applied to evaluate the performance of the prediction model. Results The decision tree, including nine independent variables and 25 nodes, was produced through the C5.0 method. The performance of the prediction model was an accuracy of 97.67%, a sensitivity of 82.52%, and a specificity of 99.15%. Fourteen classification rules were generated from the decision tree. Conclusion Significant results from this analysis were: (1) there was a specificity of 99.15%, which was 8.02% higher than 91.13% without using the decision tree; (2) prosthodontists can predict results of surgery based on a patient's physical status and implant characteristics by classification rules generated from the decision tree; (3) the original 23 independent variables were reduced to nine variables through the C5.0 method, which will allow clinical doctors to concentrate resources on fewer factors; and (4) the study showed that the variable of bone density was the most important factor (with a variable importance of 0.55) that affected the surgical results.