Research On Interactive Recommendation Algorithm Based On Perceptual Context

With the popularity of digital network information,the information on the network carrying more and more.The role of the recommended system is to allow users to get the information they need in the massive data,so a variety of recommended systems came into being,and now most of the research is how to improve the accuracy of these recommended algorithms.But as we are familiar with the situation,now the needs of different users is different,and now people are exposed to different social affairs,each person in different time nodes,places,circumstances have different preferences.This is what we want to introduce,different users have different context information,and the user's context information is often invisible to the recommendation system,which requires our recommendation system to detect the user's context,and in accordance with the the context of the context of the search for real-time recommendation,but now the context of the recommended system for the perception of the function is still relatively low,such as the process of perceiving the interaction,or how to detect the definition of these context changes,such as the need to do a lot of problems A deeper study.In this paper,the main research work on the application of the algorithm of interactive perception context is as follows:1.A context-based perceptual recommendation algorithm based on TP-Learning heuristic learning is proposed.First clear the user's context factors can greatly affect the practicality of the recommended algorithm.In many cases,especially the user's context information is changing circumstances,it is difficult to directly identify the user's preference model and the user's current context in which the user can interact with the system to detect the user At the end of the information,and dynamically determine the current user's preferences.In this paper,an interactive recommendation system is proposed to solve the dynamic stochastic optimization problem encountered in the detection process according to the user's current behavior,to detect and adapt to the user's context(or context mode)changes,heuristic learning algorithm,And for the user to get a maximum efficiency of the preference model.According to the TP-learning algorithm,the system combines the feedback information in each interaction to form a recommendation and updates the user model after the interaction.To generate a context recommendation,the user & apos;s preference model is used to monitor each change at the time of interaction with the user and to perform self-increment updates based on these changes.This is a real-time mechanism to discover changes in the user's meaningful preferences,and to make this mechanism improve the performance of the recommended system.2.Improved bipartite network structure recommendation algorithm(NBI)proposed based on time attenuation and user similarity weight recommendation algorithm(TUserCF).The recommendation algorithm based on bipartite graph is to allocate the resource values given by each user node to neighboring nodes.This paper thinks that the time factors of user's choice should be taken into account.Based on the familiarity of users and users of social network Level into account.In the process of resource allocation between nodes,the resources are not evenly distributed to the neighboring nodes,but according to the user to select the object time and user and user familiarity with the distribution coefficient to be adjusted,and finally will have more resource values and have Timeliness of the higher rating of the object priority included in the recommended list recommended to the user.Our algorithm is experimentally verified to significantly improve the average accuracy and reduce the recall rate.Our algorithm significantly improves the accuracy of the recommended items,makes recommendations more efficient,and therefore has a strong application value.3.This paper designs an e-commerce recommendation system framework,which is a combination of user specific attributes,user behavior and the third chapter of our proposed context information attributes,the structure of the N-layer Cartesian property model,the use of logical regression theory The context of the characteristics of the integration of e-commerce site recommended system architecture.This architecture is a summary of the full text recommendation system structure,is a combination of e-commerce site features and recommended system features a comprehensive product.