Research On The Algorithm Of Personalized Learning System Based On Multi-armed Bandit

As a powerful way to solve information overload,recommendation systems have been around for a long time.With the increasing scale of network users,more and more scholars have devoted themselves to the research of recommendation algorithms on various recommendation system platforms.Although it has been effectively applied in many fields,there are still some problems,such as the most common cold start problems and data sparse problems.As an emerging recommendation system,online education platform is widely welcomed by users due to its flexibility and convenience.Therefore,it has great development potential and research value.This thesis first explains the research significance of the subject and the current research status related to recommendation systems and online education platforms,and gives a detailed introduction to basic theories such as reinforcement learning,collaborative filtering algorithms,and several common multi-armed gambling machine algorithms.It also introduces Recommend the common evaluation indicators of the system.Aiming at the user's cold start problem,the main work of this paper is as follows: First,a collaborative recommendation algorithm based on multi-arm gambling machines is described.Because the traditional multi-armed gambling machine algorithm does not consider the importance of the user's feedback information to the item,and does not consider the background information and the importance of collaborative work between users in the recommendation process.Therefore,on the basis of the contextual multi-armed gambling machine algorithm,the introduction of user-based collaborative recommendation helps to improve the recommendation performance.When recommending items for a target user,the target user and neighbor users will jointly affect the recommendation results,achieving the target user's own characteristics play a leading role while the neighbor users play a collaborative recommendation role,thereby achieving the improvement of recommendation performance.After the recommendation is completed,the user characteristics are updated according to the user's real feedback and the characteristics of the recommended item,and the project characteristics are quickly fitted to the user's preferences in the few rounds as possible,effectively alleviating the impact of the user's cold start problem;secondly Based on the B2 C business model,using Java,SSM framework set,HTML,CSS and other front-end and back-end technologies to achieve a B/S architecture online education platform,and demonstrate the online education platform needs analysis,module design and other construction process and specific implementation As a result,a platform for users to obtain educational resources is provided.