Design And Implementation Of Online Education Recommendation System Based On Mixed Collaborative Filtering

Internet services in various fields are booming,and educational informatization is attracting more and more attention.Learners face the huge complexity of online education resources,and personalized recommendation system is developed for different users' needs.It not only improves the efficiency of learners,but also plays a positive role in promoting online education.Through in-depth investigation,it is found that some online education enterprises do not use recommendation system,but they recommend the most appropriate resources to users through association rules.After modification,the rules of association are too long and complex to maintain,so that rules may conflict before and after.Based on the above background,this paper discusses the necessity of education information,studies and draws lessons from the outstanding proposals made by domestic and foreign scholars on online education resources recommendation,and studies the relevant recommendation algorithms.In view of the two difficult problems of cold start and data sparsity in collaborative filtering recommendation algorithm,this paper proposes an improvement on the basis of the traditional collaborative filtering algorithm on the basis of online education scene,so that the recommendation effect is greatly improved in this scene.The main work of this paper is as follows:First,this paper selects the Pearson correlation coefficient as the similarity calculation model.In view of the problem that the user-project score matrix sparsity leads to the reduction of the recommended quality,a method of improving the time attenuation function and the weight of the number of common score items is proposed.The basic idea is: first,the time attenuation function is defined.If the time interval between the two users is smaller,the higher the similarity between the users;second,the weight of the common score between the users is incorporated into the calculation of the similarity.Highlight the contribution of users who share a large number of ratings in computing similarity.Second,on the basis of the improved Pearson coefficient calculation model,this paper proposes a method of combining user characteristics to find the nearest neighbor set to solve the problem of user cold start.Reasonable design of dynamic harmonic parameters can effectively combine user attribute attribute similarity with Pearson coefficient similarity.When new users enter the system,they rely solely on the user's basic information to assess user similarity.This is a more reasonable solution.Third,combined with the real data set of an online education company,the modified hybrid collaborative filtering recommendation algorithm is compared with the traditional algorithm.It is proved that this algorithm can adjust the similarity between users through the mixed Pearson coefficient model and user characteristics.On the one hand,it alleviates the cold start problem of the new student into the system to a certain extent;on the other hand,it improves the credibility of the recommendation system when the data is sparse.The modified hybrid collaborative filtering recommendation algorithm has more effective and higher quality recommendation than the traditional algorithm in Personalized Education Resource Recommendation scenario.Based on the collaborative collaborative filtering recommendation algorithm,an online education recommendation system is developed for supporting the system back stage recommendation engine.This is a dynamic system that accepts users' ratings in real time and presents them to users.When user behavior information is more sufficient,the recommendation results will be more convergent.