Research On Learning Recommendation Algorithm In Collaborative Learning Environment

Continually, the development of e-commerce has promoted progressing of recommendationfield, escalating of recommendation system and improving of recommendation algorithm. Also,more and more people realized recommendation was so convenient, so that more researchersstudied recommendation system. Similarly, the progress and improvement of online educationmade recommendation system played a role in the online collaborative learning platform.However, with the rapid increasing of users and resources, online collaborative learning andrecommendation algorithm have exposed its own defects. Online collaborative learning couldnot provide personalized learning and recommendation system greatly reduced its effectivenessand accuracy. That had seriously affected learning experience and learning efficiency incollaborative learning platform. Therefore, a collaborative filtering recommendation algorithmin collaborative learning environments is proposed by combining and expanding characteristicsof collaborative learning environment and collaborative filtering recommendation algorithm.The algorithm reflects the "personalized learning" characteristics in cooperative learning, butalso improves the accuracyof calculating similarityand recommendationFirstly, the important attribute values of learners and learning resources are extracted afteranalyzing characteristics of learners and learning resources in collaborative learningenvironment. Next, computer identification and processing become convenient when allattribute values map a single mathematical set according to the theory of hierarchy. Then,attributes of learners constitute learning attribute matrix and the comprehensive similarityconsists of attribute matrix similarity and user rating matrix similarity. Besides, the author alsoconsiders influence of time that recent interest is more important. Thus, a time function isapplied to the final recommendation and different similarity of users has the different weights.The improvement of the recommendation algorithm is a major feature of this paper.In the experiment, the algorithm is verified using the simulation data in the simulationplatform. In order to reflect the objectivity and comprehensiveness, the author uses the mean absolute error to measure the accuracy of the algorithm in different number of learners, differentnearest neighbor K value and different degree of sparse matrix. And the algorithm is in contrastto a successful collaborative filtering recommendation algorithm. The experimental resultshows that the proposed collaborative filtering recommendation algorithm is more preponderantand more accurate. Also, the algorithm profoundly reflects the important characteristics of"personalized learning".