Design And Implementation Of Adaptive Learning System Based On Knowledge Graph

With the advent of the Internet+ and Big Data strategy,The deep integration of Internet innovation technology and various fields has become the focus of attention,Meanwhile,it also promotes the reform and innovation of education in China.2015 is the first year of China's education about big data.A large number of online education products came into being,which has brought great convenience to people.However,facing the massive learning resources on the platform,Learners often have issues such as network trek,cognitive overload,and knowledge fragmentation.The emergence of adaptive learning concept will bring new opportunities for the future development of online education.At present,domestic online education products are mostly data statistics and test question bank types.Although individual products use relatively complex logic derivation,However,it is still far from the real adaptive learning.Research on adaptive learning is mostly focused on the theoretical level,Concerns about the realization of system technology are still far from enough.This paper is from the perspective of learners,Fully consider the structural features of domain knowledge,Focus on implementing a series of key technologies of big data in adaptive learning systems,which hopes to provide reference for the research and application of education big data.The main work of this paper is as follows:(1)Based on analysis of functional requirements of the adaptive learning system.The overall framework of the system,workflow and the construction method of each unit module are designed.(2)Based on the association characteristic of knowledge graphs,the knowledge representation of domain models is completed,and the construction methods and application examples of knowledge graphs are given.On this basis,using the fp-growth algorithm to mine the association rules of student path data.And to complete the dynamic update of the domain model.(3)The construction method,construction process and storage structure design of the student model are completed by referring to the student model construction standard of our country.Based on the IRT test model,the test of students' cognitive level and the construction of test question bank were completed.(4)Based on the SVR predictive regression model to achieve the prediction of path completion,And completed the SVR prediction model training,parameter tuning and forecasting.The experimental results show:(1)in the domain model,the knowledge graph can achieve the function of clear expression of the knowledge structure.FP-Growth algorithm can realize the mining of association rules.The minimum support degree was determined as 0.03 by contrast analysis;(2)In the student model,the parameter estimates given by the IRT test model conform to normal learning patterns and have practicality;(3)In the adaptive engine module,the optimal combination of the four parameters of the SVR prediction model is determined: The non-sensitivity coefficient is 0.05,the penalty coefficient is 8,the kernel function is RBF and the kernel parameter is 0.08,and the system prediction error is about 0.04.The main contributions and innovations of this paper are as follows:(1)Starting from the domain knowledge structure,using the knowledge graph to complete the effective expression of the knowledge structure,and put forward new research tools for the construction of the domain model.Using association rule mining algorithms to dynamically update the model,Fully consider the influence of student factors on the domain model and improve the practicality of the model.(2)This paper proposes a predictive regression model based on SVR to implement the adaptive engine function.Integrate machine learning ideas with education,It is an exploration and innovation of the study of adaptive learning system,and provides new research ideas for related researchers and developers.(3)This paper proposes a test model based on IRT to realize the evaluation function of students' knowledge mastery,distinguishes students' ability under the same score and improves the objectivity and accuracy of the student model.