Design And Implementation Of A Learning Resource Recommendation System Based On Stream Computing Platform

In recent years,educational informatization has made a transition to educational innovation.The recognition and utilization rate of online education has increased significantly,and online education platforms are also constantly exploring,combining advanced technologies to transform education methods.In order to ensure that students can obtain high-quality learning resources in time when using teaching auxiliary products,this thesis aims to implement a system that can recommend relevant learning resources in real time according to students’ current learning status and historical learning data,so as to improve the pertinence of learning content and effectively improve the learning effect of students.This thesis proposes and builds a deep learning-based learning resource recommendation model.Based on the architecture of Deep FM,an attention mechanism is introduced to design a new model structure.The Item2vec method is used to perform Embedding processing on sparse features,and attention weights are calculated according to the similarity between candidate resources and user historical learning resources,and the processed features are input into the model.The FM part and the DNN part are added to output the prediction result.In the laboratory environment,several recommended methods of the investigation are tested and compared with offline indicators,and the feasibility of the model is proved by the experimental results.This thesis builds a complete data closed-loop recommendation system based on the model.After getting the trained model,put the model online and build the server.Using stream computing processing platform to process real-time data.Through the processing of real-time streaming data,the user’s Embedding vector is dynamically updated,input into the recommendation model,and finally time-sensitive recommendations are displayed for the user.At the same time,this thesis uses the batch data processing platform to extract features and train the network.As the offline data processing part,the real-time data processing part and the recommendation model part together constitute the whole system.Finally,the system was applied to the online teaching assistance platform,and A/B tests were performed on users in the online environment.The online indicators of the experimental group were better than those of the control group,indicating that the system built in this thesis showed a good recommendation effect and can be used in application in the actual engineering environment.