Research On The Application Of Data Mining In Student Online Testing And Prediction

Education is a decisive cause for the great rejuvenation of the Chinese nation.Educational data mining is a new and multi-disciplinary research field,exploring the methods and techniques to obtain data from various educational information systems.Educational data mining has been widely recognized by the researchers in the fields of machine learning,artificial intelligence,education,cognition and so on.This paper studies the learner's knowledge model of the knowledge mastery state.It is the core research field in the education data mining.It has good research value and practical significance.This article mainly completed the following three aspects of the work.Firstly,it studies how to construct the learner's knowledge model,and to understand and master the students' learning process and mastery of knowledge by digging the log data of the intelligent teaching system.Second,it compares the prediction and real results of the model,gets the evaluation index,and evaluates the degree of conformity between the learner's knowledge model and the actual situation.Third,it applies the research results to the actual data obtained from the school education system,proving that the model obtained by this method has certain accuracy.Specifically,the paper takes the open data set of the KDD CUP 2010 competition as the research object,uses a variety of preprocessing methods,and reduces the data of the size of the 5.29 G and the 20 million samples into the data set that the personal computer can excavate through the tedious and extremely time-consuming processing,and then carries out modeling and model evaluation.The accuracy of this model is 88.9495%,the root mean square error RMSE is 0.2848,which is a little worse than the RMSE 0.271157 of the competition champion.The champion is the 25 Deluxe team,led by Lin Zhiren,a famous professor at National Taiwan University.Compared with the models constructed by logistic regression,SVM algorithm,Bayesian network and BP neural network,we find that the model is the best.This paper also evaluates the generalization ability of the model,which proves that the model has good generalization ability in extreme cases.Finally,the actual data of the students in the online education platform are excavated,and the accuracy rate is 82.3301%.This study follows the famous Occam razor principle in machine learning: the simplest explanation of data is the best explanation.The 20 features in the open data set are converted to 3 features by preprocessing,which greatly simplifies the problem and achieves good results.