Personalized real-time recommendation method for online learning resources

Abstract

The invention provides a personalized real-time recommendation method for online learning resources. According to the knowledge points selected and self-cognitive ability input by a learner in an online learning system through an input display unit of a user terminal, learning resources are recommended to the learner in real time in a personalized manner. The method is characterized by comprisingthe following steps: in step 1: the learner's learning feature parameters and course feature parameters of a recommended course are obtained from the online learning system; in step 2, a learner modeland a learning resource model are built; in step 3: an objective function with an optimal recommendation strategy is determined; in step 4, a learning resource storage structure map with an orthogonal list structure is built; in step 5, key parameters of the objective function are determined; in step 6, a binary differential evolutionary algorithm is adopted to obtain an optimal selection combination of learning resources; in step 7, the learning resources are separately sent to the learner's user terminal according to the best selection combination, and therefore the learning resources are recommended to the learner.

Description

technical field [0001] The invention belongs to the field of electronic teaching, and in particular relates to a method for personalized real-time recommendation of network online learning resources. Background technique [0002] The Ministry of Education continues to encourage colleges and universities to build and open high-level online open courses to form a large number of open and shared high-quality educational digital learning resources with a complete range of subjects. Usually learners will not know where to start when faced with massive learning resources, and are full of confusion. They can only try to find learning resources suitable for their own learning progress in the form of continuous trial listening and trial viewing, which makes this process time-consuming and labor-intensive. The results are also unsatisfactory, delaying the learning progress of the learners. Therefore, the personalized recommendation technology of learning resources is an effective tec...

AI Technical Summary

Problems solved by technology

[0004] In terms of user modeling, foreign related research includes: In 2014, Rtili, M.K modeled the learners by collecting and processing the learner's interaction traces, so that a group of interacting agents could automatically propose the appropriate user information for them. Educational resources, but the recommendation accuracy is not high (Journal of Emerging Technologies in WebIntelligence, 2014, 6(03): 340-347.)

In 2017, Tarus, J.K proposed a heterogeneous educational resource recommendation system based on user preferences based on user needs. The system is efficient but the recommendation quality is not high (Future Generation Computer Systems, 2017, 72: 37-48.)

Domestic research includes: In 2013, Tu Jinlong used newly opened tabs to build a user (learner) interest model, so that the recommendation of learning resources would change with the change of user interest, but the timeliness of this recommendation method is not high (computer Applied Research, 2013,30(04):1044-1047+1054.)

In 2017, Zhang Xiaoxue used learner self-assessment combined with the Felder-Silverman scale to analyze the learner's learning characteristics, so as to build a learner's relevant learning model and make targeted learning recommendations. However, this method not only requires learners The construction of a personalized recommendation service model for online learning resources still needs to spend a lot of time on learning self-assessment (Chinese Medical Education Technology, 2017,31(02):172-176.)

[0005] In terms of content research, foreign related research includes: In 2013, Salehi, M. used the attributes of learning resources and the sequence mode of learners' access to resources in the recommendation process to recommend learning resources. This method introduced the learning tree (LT), Considering the explicit multi-attributes of resources, the recommendation accuracy is high, but the knowledge base of the learners is not considered, and the recommended resources cannot match the learners' acceptance ability (Data and Knowledge Engineering, 2013, 87:130-145.)

In 2016, Alinani and Karim used ontology domain knowledge and learning sequential access pattern mining system to recommend learning resources, but the recommendation accuracy was not high (International Journal of Autonomous and Adaptive Communications Systems, 2016,9(02):20-39 .)

Domestic: In 2014, Xu Shoukun constructed the learning resource usage ontology and used semantic reasoning to enrich the resource recommendation result set, but it was difficult to accurately construct and initialize a multi-semantic association learning resource (Computer Engineering and Design, 2014, 35( 04):1496-1501.)

In 2016, Liu Meng deeply analyzed the knowledge points of professional courses for professional learning, and used reachability matrix and parallel topological sorting methods to recommend learning resource paths. For online resource recommendation (Computer Simulation, 2016,33(06):180-184.)

[0006] In terms of recommendation strategy: In 2014, Zhang Haidong used association rule mining and similarity methods to determine the association between any courses or resources, and recommended learning resources to primary and middle school students, but the accuracy of the recommendation is difficult to guarantee (Computer Applications, 2014, 34(11):3353-3356+3364.)

In 2016, Cheng Chunlei established a relational concept semantic identification model based on the knowledge relational concept as the semantic basic unit, which is used for the recommendation of web personalized learning resources, but this method is difficult to realize the dynamic adjustment of the model and parameters (Computer Science and Exploration, 2016 ,10(08):1092-1103.)

In this regard, Zhu Xia built a personalized learning resource recommendation method based on collaborative filtering. This method has high recommendation accuracy and efficiency, but it cannot solve the cold start problem because of relying on the historical records of learning behavior (Computer Research and Development, 2014, 51( 10):2255-2269.)

In 2017, Khosravi, H used a matrix factorization-based collaborative filtering algorithm to provide individual students with personalized recommendations to address their interests and current knowledge gaps, but the algorithm is not universal (arXiv, 2017, 25.)

The recommendation of learning resources is transformed into a multi-objective optimization problem, and then the particle swarm optimization algorithm is used to solve it to form the optimal recommendation strategy. However, the selection range of learning objectives cannot be dynamically adjusted and the algorithm complexity is high, which is not suitable for online recommendation (computer application ,2014,34(05):1350-1353.)

[0007] Therefore, the above-mentioned existing recommendation strategies have the defects of difficult dynamic adjustment of learning objectives and poor real-time performance.

Images