Affected by the COVID-19 epidemic,offline teaching activities across the country were suspended,and online education(E-Learning)ushered in an opportunity for development.Users can use a computer or mobile device to learn online,so that learning is no longer limited by time and space.However,with the strong demand for the online education market,major platforms need to surpass their peers in terms of number and quality of courses to attract more users.As a result,the number of courses on the online education platform has increased exponentially year by year,and the problem of course overload is serious.It is difficult for users to quickly search for interesting courses from the massive courses.An effective way to solve the overload of course information is to apply the recommendation system to the online education platform.By collecting and modeling the user's learning behavior data on the platform,and then mining and analyzing the constructed recommendation model,you can predict that the user may be interested.Courses to provide users with personalized course recommendations.The main research contents of this article are as follows:(1)Through the demand analysis and design of the E-Learning platform curriculum recommendation system prototype,the user use cases,functional structure and overall system architecture of the system prototype are determined,and the recommendation system recommendation process and database are designed in detail.(2)The basic principles,advantages and disadvantages of the collaborative filtering recommendation algorithm based on ALS(Alternating Least Square)and the content recommendation algorithm based on TF-IDF(Term Frequency-Inverse Document Frequency)Research and analyze the feasibility of mixing the above two algorithms.In the end,this paper adopts the feature supplementary hybrid strategy to design the hybrid recommendation algorithm.First,the feature weight calculation module in the content recommendation algorithm based on TF-IDF is improved,and the time factor is integrated to make the algorithm more suitable for the knowledge characteristics of the course.Then,the recommendation results based on the content recommendation algorithm are used to generate a virtual user course scoring matrix and the virtual course The scoring matrix is filled into the original course scoring matrix,and then the ALS collaborative filtering algorithm is used to make scoring predictions to obtain the final recommendation result.Finally,in order to verify the effectiveness of the hybrid recommendation model,this paper conducts an experimental analysis on the prediction accuracy and classification accuracy of the recommendation algorithm.After experimental verification,the hybrid recommendation method designed in this paper has a better recommendation effect than the traditional collaborative filtering algorithm.(3)A course recommendation system is implemented on the Spark distributed computing platform,and the computing power of the system is improved with the help of Spark clusters.In order to improve the real-time nature of the course recommendation system,this paper implements both real-time and offline recommendation services.Based on the prediction results obtained by the offline model,Kafka and Spark Streaming are used for real-time recommendation,which reduces the time for users to obtain recommended courses. |