Cheating Detection Method Based on Improved Cognitive Diagnosis Model

Zhizhuang Li, Zhengzhou Zhu, Teng Yang

Cheating in examinations destroys the principles of fairness and justice in evaluation. Cheating detection is of great practical significance. Traditional cheating detection methods have many disadvantages, such as difficult to detect covert equipment cheating, multi-source cheating, difficult to distinguish plagiarists from plagiarists, difficult to distinguish plagiarists from victims, or plagiarism from coincidences. In this paper, the concept of knowledge point mastery Index is introduced to measure students‘ mastery of a certain knowledge point, and a test method of cheating based on improved cognitive diagnostic model is proposed. This method calculates the weight of each knowledge point in every examination question through linear regression and EM algorithm according to students‘ historical learning behavior, and then calculates students‘ mastering degree of knowledge point based on historical answers. Then calculate the mastering degree of knowledge point based on the examination results. Finally, we compare the mastering degree of knowledge point based on the examination results and the historical answers to detect students‘ cheating situation. The experiments show that the average precision ratio of this method is 31.3% higher than that of the method based on the false-same rate, 23.9% higher than that of the method based on the false-same rate and the right-same ratio, 96.2% higher than that of the method based on the Person-Fit index. The recall ratio is 74.8% higher than that based on the false-same rate, 114.0% higher than that based on the false-same rate and the right-same ratio, and 27.8% higher than that based on the Person-Fit index.