The purpose of this research is to analysis how to establish a computerized diagnostic test system that can be used to identify the error patterns of a student by Bayesian network and is applicable to make a remedial online teaching. It is very important to improve students’ wrong concept in algebra after computer remedial instruction. Therefore, the research is to seek for further understanding and to improve student’s key competencies. The assessment task of error patterns was administered to 66 sixth-grade students from Taichung and Nantou County. The key points are described as follows： 1. A study of using evidence process to construct a concept evaluation model, the related contents can be applied to diagnosis error patterns of students. 2. The study found that the most causes of the error patterns were the conception of 「4 Basic Arithmatic Operations」in first diagnostic test. After computer remedial instruction, the error patterns of 「Algorithm」was reduced obviously. The results indicated that the student exhibits a substantial increase in the ability index. 3. The mistake was evidently improved for high-score group students in second diagnostic test, but there was no significant improved for low-score group students. Results suggest that the computer remedial instruction was affected those who have better learning ability according to the high-score group part. 4. Data analysis is conducted by the utilization of t-test, the findings generated that the computer remedial instruction was exhibited a significant effect on improving error patterns and increasing score. 5. Based on the Bayesian Network, there is high level of trustworthiness and applicability using the computer diagnostic assessment for six-grade student’s subject of algebra. We can find the better learning method for mathematics concepts in the future.
