Study on student performance estimation, student progress analysis, and student potential prediction based on data mining.

Abstract

Student performance, student progress and student potential are critical for measuring learning results, selecting learning materials and learning activities. However, existing work doesn't provide enough analysis tools to analyze how students performed, which factors would affect their performance, in which way students can make progress, and whether students have potential to perform better. To solve those problems, we have provided multiple analysis tools to analyze student performance, student progress and student potentials in different ways. First, this paper formulates student model with performance related attributes and non-performance related attributes by Student Attribute Matrix (SAM), which quantifies student attributes, so that we can use it to make further analysis. Second, this paper provides a student performance estimation tools using Back Propagation Neural Network (BP-NN) based on classification, which can estimate student performance/attributes according to students' prior knowledge as well as the performance/attributes of other students who have similar characteristics. Third, this paper proposes student progress indicators and attribute causal relationship predicator based on BP-NN to comprehensively describe student progress on various aspects together with their causal relationships. Those indicators and predicator can tell how much a factor would affect student performance, so that we can train up students on purpose. Finally, this paper proposes a student potential function that evaluates student achievement and development of such attributes. We have illustrated our analysis tools by using real academic performance data collected from 60 high school students. Evaluation results show that the proposed tools can give correct and more accurate results, and also offer a better understanding on student progress.