Predicting Students’ Online Grades Based On Data Mining Technology

In the post-epidemic era,offline education has suffered a lot in varying degrees.The popularity of the Internet and the development of artificial intelligence technology is profoundly changing the direction of education innovation,breaking the barriers between online and offline,and making online education models emerge.With the changes in education methods,how to evaluate the effectiveness of online education has become an urgent problem to be solved.Students will accumulate a large amount of data throughout the learning process,and it is a good solution to explore the laws of association between these data through data mining methods.As one of the popular research fields of educational data mining,test score prediction is to predict the learning status and learning effect of students.The continuous expansion of the scale of online education,the continuous renovation of the format,and the flexible and changeable teaching methods make it difficult for education managers to obtain teaching feedback opportunely and carry out educational interventions.More than this,it will also affect the development of students and the development ecology of online education.Exploring and constructing efficient and accurate performance prediction methods has become a research direction with important application value and significance.This thesis takes the online academic education model of a provincial open university in Shaanxi as the starting point.In the theoretical research stage,through literature analysis,it combines classic theories and models from three aspects,including social characteristics group attributes,individual attributes,and learning behaviors.The status quo of domestic and foreign research on performance prediction is reviewed.By analyzing the data of more than 80,000 students in the school,researching and predicting students’ final exam results,an improved clustering algorithm model and two prediction models are proposed.The main content and innovations of this thesis are as follows:First of all,by organizing,aggregating and preprocessing data from multiple sources about the school’s online academic education students,behavior and attribute characteristics of students are analyzed.By extracting effective behavior characteristics and combining the actual situation of online academic education which can optimize the K-means algorithm and perform clustering,the accuracy and effectiveness of clustering are further improved.Secondly,by using the work of the preliminary data analysis stage,the influencing factors that affect the final exam scores are determined.And the optimized clustering center is used to make it the center of the neural network,and then the RBF neural network is optimized as a performance prediction model.Accuracy and effectiveness of the optimized RBF neural network prediction model can be proved by all the above experiments.At last,based on previous course,the method of "pre-semester" for predicting the final exam scores is put forward.This thesis mainly uses some "past" course information that students have completed before to predict the final exam scores of the new semester,which effectively takes advantage of the correlation between different courses and the students’ previous learning status and exam scores.Multi-instance and multi-label learning methods are introduced to make predictions.This method and the RBF prediction model introduced in this article complement each other.It can predict the course performance before the start of the course,so that the timeliness of the performance prediction is advanced and it is more predictable.