Track: High School STEM Project Presentation

Abstract

This Poster will demonstrate on how to apply Big Data and Data Mining Analytics on detecting the cheating pattern.  JMP Statistics software was utilized to analyze the multi-correlation among 75 students (sit on 25 different tables) taking an assessment exam questions with multiple choices.  Three students from the same table will take the same exam but with different sequence of questions.  However, students were smart to synchronize the question sequence immediately.  In order to detect the cheating pattern, multivariate statistics was used to determine whether there was any association among the students from the same table.  Hierarchical Clustering and Dendrogram Tree can identify the grouping affinity behavior related to exam cheating pattern.  Team also used graphical Heat Map to visualize the cheating pattern.   In order to improve the prediction confidence, team also selects the top 20% difficult questions to increase the detection signal-noise ratio to minimize both Alpha and Beta Error below 1%.  At such lower level, students won’t defend their cases statistically.  Based on the results, three tables were identified with cheating patterns.  The predictive model based on the Big Data Analytics was very powerful to analyze complicated cheating patterns.