Practice of Data Mining in Formative Evaluation

Abstract

The main purpose of this paper is to use the students' English learning situation on the Internet to formally evaluate the students' final English performance level. First of all, we introduce the concept of formative evaluation, and the principles of three kinds of data mining algorithms: naive Bayes classification, C4.5 decision tree, and Logistic regression; then, we use the student online learning data table to achieve the key calculation process of the above algorithm; Further, we use Matlab programming to predict the student's final grade level and compare the performance of each algorithm. Practice shows that, C4.5 performs better than Naive Bayes algorithm on predicting the four classifications of grades (great/good/medium/bad), but the accuracy is not very high; Naive Bayes performs better than the other two algorithms and has higher accuracy on predicting the two classifications of grades (good/bad). Considering the two factors of duration of online learning and number of submissions, the accuracy of the prediction has not been significantly improved. Therefore, there is no need to consider both in terms of this formative assessment. Formative assessment has a very important significance in teaching, and plays a key role in motivating students' learning and teacher guidance. According to the forecast results, it can provide some help and guidance for students' follow-up study, so as to improve students' learning effect.