Research On Micro-Learning User Type Identification Based On Improved Deep Belief Network

The shortening of the knowledge renew cycle requires people to continuously renew their knowledge.The accelerating pace of work and learning has made people choose more online learning resources based on computer and internet technology,such as using Weibo,WeChat,and micro-courses as platforms,in order to learn at anywhere and anytime.However,the fragmentation of work and learning time caused by the multi-tasking of work and learning makes it difficult for learners to spare an entire time for learning.Therefore,a platform that can use a short time for learning has become a need.This need has spawned a new learning style,which is named as Micro-learning.As a new learning style,micro-learning is gradually expanding from the practice in social field to school education.Although the existing micro-learning platform can provide learning resources based on a short learning time,the learning resources show a multi-source heterogeneous feature in the internet,and the emergence of a large number of learning resources also causes problem such as information overload,which leads to more time will be spent to search suitable learning resources.For the publishers of learning resources,it is also a thorny problem to make their published learning resources to be quickly communicated to suitable learners.Therefore,classifying learners of different ages,different industries and different educational background can help to provide learners with learning resources,which has great significance for learners'personalized learning environment.In order to detect the needs of learners in the micro-learning platform,an improved deep confidence network model based on the multilayered sparse class constrained Boltzmann machine is proposed in this study,which is used to discover the types of resources that learners are interested in.This model is based on micro-learning related theories,and combined with deep belief networks,which can be used to identify the types of micro-learners through improved sparse coding models and restricted Boltzmann machine models.The main work of this paper is shown as follows:(1)A category layer is added to the restricted Boltzmann machine model to identify the type of micro-learners in the supervised learning.(2)The sparse constraint is used into the algorithm of deep confidence network model,and salient features are selected to identify the type of learners.(3)The arctan function is used to replace the penalty term of the original sparse constraint,and obtain a more effective sparse representation of the target and realize the recognition of the learner type with more concise and effective training.The comparative experiments show that the improved sparse deep confidence network can improve the accuracy of user type identification,improve the accuracy of user classification,and improve the performance of the model for user type identification.