A case study in a pediatric dental clinic was presented. The data were transformed into LRFM (Length, Recency, Frequency, and Monetary) format with fixed M covered by National Health Insurance program in Taiwan, where the data were categorized into 1 to 5 for L, R, and F variables. Later, gender was classified into two types, and age was grouped into four categories. The target in this study was frequency, while L, R, gender, and age were the input variables when Bayesian network was performed. The results show that the overall accuracy is 65.26%, and three out of five classes have relatively high accuracy values. Moreover, the value of the overall receiver operating characteristic (ROC) area is 0.891, which indicates that this Bayesian network model performs well in this pediatric dental clinic study. Furthermore, recency and age are the two better variables to forecast frequency.