Research On Adaptation Of Recommendation Algorithms In Different Learning Scenarios

Traditional learning methods are limited to the classroom,and the source of learning resources is limited to the library.In such a learning environment,every student's learning behavior is greatly limited.The emergence of campus system based on cloud desktop technology and MOOC Course has broken the time and space limitations of learning.However,because of these limitations being broken through,students face a wide variety of massive learning resources directly.How to quickly acquire learning resources that are really suitable for their own needs in these learning resources has become the primary problem that learning objects need to solve.How to make learning objects more simple and fast access to learning resources that meet their own needs is the focus of this paper.According to the current situation,this paper summarizes three kinds of learning scenarios: learning scenarios based on MOOC lessons,learning scenarios based on academic research and learning scenarios based on campus teaching.Traditional campus systems often focus on classroom teaching and campus announcements.The learning behavior of learning objects is only recorded,but these data are not fully utilized.By analyzing the characteristics of these scenarios,this paper discusses the adaptation of common recommendation algorithms in these learning scenarios.Finally,according to the characteristics of the recommendation algorithm and the actual situation of the learning scenario,the adaptation of the algorithm is studied.In the learning scenario based on MOOC,according to the specific situation of learning behavior and the feedback of learning objects,the article-based collaborative filtering algorithm is adapted.The attributes of learning resources model is established.The first 80% of the data parsed by the imooc log files are trained to get the recommendation list,and then the recommendation accuracy is obtained by comparing the last 20% of the data.The accuracy rate of the proposed algorithm is 25.6%,which is 41.4% higher than that of the original algorithm in the imooc website.In learning scenarios based on academic research,through the analysis of learning scenarios and the induction of the characteristics of learning resources,a content-based recommendation algorithm in learning scenarios is proposed.Collecting the time spent by some learners in the process of writing academic papers and the number of academic papers actually consulted as the experimental data of the algorithm,it proves that this algorithm can help learners to reduce the number of papers consulted by 31.7% and the consulting time by 33.2% in academic research.In the teaching-based learning scenario,according to the actual use of the campus cloud desktop,the personal characteristics and learning characteristics of the learning objects that can be collected are analyzed.Because of the complexity of these learning data,this paper adopts hybrid recommendation algorithm in this learning scenario.According to different weights of content-based recommendation algorithm,item-based collaborative filtering algorithm and user-based collaborative filtering algorithm,a complete process of hybrid recommendation algorithm is designed.The simulation data are obtained by questionnaire,and the hybrid algorithm is evaluated.The weight ratio of the ideal hybrid model algorithm is obtained.The final recommendation success rate of the hybrid algorithm is 32.676%.