Bayesian Networks is a probability analysis method in medicine and industrial engineering; it is also applied in education recently. The main idea of the study is to research the ability of the geometrical design in elementary and demonstrate the applicable of students mistaken types based on Bayesian Networks which is a probability analysis method on the basis of the ontology. Students join the test through the Learning Educational Program online, which is developed based on the index of Grade 1-9 Curriculum in Taiwan. The system can show the subject comprehension and begin to the Learning Educational Program on the basis of the mistaken type distribution. There are three main purposes of the study as below. 1. Diagnosing the geometiy index of the mistaken types in elementary by using Bayesian Networks on the basis of the ontology. 2. Establishing the Computerized Remedy Instructions and the Learning Educational Program based on the study model. 3. Demonstrating the effective of the computerized adaptive flash. There are two main conclusions of the study .According to the analysis of the test result by Bayesian Networks Model, the average corrective is up to 90% with mistaken types and up to 80% with sub-skills and capable index. It shows that the practicability of Bayesian Networks is applied on the concept of the geometrical design in elementary. It shows the different results after giving the suitable teaching to the students by the attracting flash, and then the students get higher scores than before. (t-test, p
