Study On Evaluation Model Of Adolescent Learning Literacy

Mc Kinsey first proposed the arrival of the era of big data.Nowadays,big data is widely and effectively used in the fields of biology,medicine,online finance and communication.It is inevitable that big data also brings unprecedented challenges and opportunities to China's education reform and development.At the same time,education data mining technology(EDM)was born,it used to find and solve problems in the field of education.It is worth further study that How to effectively use EDM technology to construct an education-related evaluation model.Educational evaluation is divided into educational measurement and educational evaluation.Educational evaluation is an important way to correctly understand and evaluate the essential law of educational phenomena,and it is also a key link affecting the educational reform and development in China.Based on the questionnaire data of PISA2018 published by OECD,this article explores the relationship between student characteristic variables and students' literacy scores in reading,mathematics and science,with the purpose of building an evaluation model of adolescent learning literacy based on EDM.The construction of educational evaluation model is divided into two parts.In the first part,the fitting regression model was used to predict the learning literacy scores of adolescents,and all the data were extracted by means of k-fold cross validation.By comparing SVR,KNN,BP neural network and RF algorithm,it is found that SVR performs better in the test set,the goodness of fit can reach 83.7%.By further comparing the linear SVR with the improved linear SVR,it is found that the performance of LSSVR is better than that of ordinary linear SVR,and the importance of each influence factor can be observed according to the weight of the improved linear SVR.In the second part,the model is constructed to distinguish the poor and excellent learning literacy of teenagers.Four kinds of algorithms are also tested.Firstly,double crossover verification is carried out for each algorithm.On the one hand,k-fold crossover verification is carried out for the original data set.on the other hand,the optimal parameter selection of the model is cross-verified,so as to select the optimal parameter model of each algorithm.Then,a horizontal comparison of the four algorithms shows that SVM and RF have better performance.In addition,according to RF's ranking of the importance of input variables,it is found that students' metacognitive ability,learning duration,age at the beginning of compulsory education,parents' education level and parents' occupational status index have a great influence on teenagers' learning literacy.Finally,the evaluation model of adolescent learning literacy is constructed.At the same time,it is found that students' metacognitive ability,learning duration,age at the beginning of compulsory education,parents' education level and parents' occupational status index have great influence on students' learning literacy.It provides scientific and effective improvement methods for the stakeholders in the field of education.