Font Size: a A A

The Research On Personalized Learning Materials Recommendation Based On Hybrid Differential Evolution Algorithm

Posted on:2019-08-06Degree:MasterType:Thesis
Country:ChinaCandidate:G ZhangFull Text:PDF
GTID:2417330596964760Subject:Education Technology
Abstract/Summary:PDF Full Text Request
With the application of artificial intelligence technology and the construction of online learning systems,the learning resources are becoming more and more abundant,and it is more and more difficult for learners to select their own learning resources from a large number of learning resources.For the intelligent learning system,providing personalized recommendation services for learners has always been the core function of the intelligent learning system.However,collaborative filtering recommendation algorithms widely used in e-commerce and news recommendation fields,the learning resources recommended by which do not have stronger inherent logic,and do not meet the learning rule of building knowledge based on prior knowledge.Therefore,it is not entirely applicable to learning resources recommended application scenario.In recent years,scholars both at home and abroad have defined the learning resource recommendation as a multi-objective combinatorial optimization problem.Discrete particle swarm optimization and other evolutionary algorithms are used to solve the multi-objective combinatorial problem of learning resource recommendation.However,the algorithm needs too much advance testing to predict the model parameters,which is not in line with the trend of online intelligent learning.In view of the shortcomings in the current field of recommended learning resources,firstly,a new portrait of learner portraits and learning resources is established.Based on the portrait model,a new learning resource recommendation model is proposed.The recommended model establishes a model for individual learners,and Learning resources are divided into knowledge points,each time a learning resource is recommended to a single learner based on knowledge points.Secondly,for the problem that too many parameters in the recommendation model need to be tested in advance,and it is difficult to accurately estimate,a user-based collaborative filtering algorithm is used to estimate and find the learner most similar to the current learner,Because similar learners have a consistent level of knowledge,so learning data of similar learners is used to estimate the learner’s master of new knowledge points,interest,enthusiasm and other parameters.Thirdly,an improved differential evolution algorithm is proposed,which enhance the basic differential evolution algorithm and improve the shortcomings of the algorithm’s that have low recommend accuracy and is easy to fall into the local optimum problem.Finally,a new hybrid recommendation algorithm is proposed,which combines the collaborative filtering algorithm with the discrete differential evolution algorithm to solve the specific complex problem of learning resource recommendation.In order to verify the model and algorithm effectively,a detailed experiment is designed to verify the algorithm.Firstly,the collaborative filtering algorithm and other binary particle swarm algorithms are merged.Secondly,the convergence,convergence accuracy and execution time of the algorithm are compared and analyzed under different conditions of the recommended number of learning resources.Finally,according to the operation of different algorithms,the causes of the differences in the operation of the algorithm are compared and analyzed in detail.Through the convergence experiment,the feasibility of the algorithm is verified.Through the comparison and analysis between different algorithms,it is proved that the algorithm is superior to the other two algorithms in convergence accuracy and execution time.
Keywords/Search Tags:personalized learning resource recommendations, learner portrait, learning resource recommendation model, collaborative filtering recommendation algorithm, differential evolution algorithm, hybrid learning resource recommendation algorithm
PDF Full Text Request
Related items