The Research Of E-Commerce Personalized Recommender Systems

Facing the new tides of globalization and digitalization, hidden troubles have increasingly emerged in the publishing industry which should have an even broader market than ever before. As a new business model, not only can e-commerce change the laggard management situation of the publishing industry for the better, but also open a new market strategically.The e-commerce personalized recommender systems, by applying the knowledge discovery technology, can make personalized recommendations on products and services to different users. The application of recommender systems will definitely redound to enhance e-commerce competitiveness in the industry.The thesis has made an in-depth study on e-commerce personalized recommender systems, analyzing the problems in collaborative filtering recommender system in detail. With these foundations, it continues to put forward a modified collaborative filtering recommender system supported by an assistant item rating segment and a rating prompting mechanism.The modified recommender system is strengthened by an assistant item rating segment based on the original one. The user-item rating database can be upgraded by supplementing it with assistant items ratings information. The similarities between each item are computed offline according to comparability measurement. Under the theory of item-based collaborative filtering arithmetic, the system predicts and makes up the ratings of items which haven't been rated. On the basis of the predicted results, the system computes the correlation of users and finds the neighbors of target user. With an assistant item rating segment, the original collaborative filtering recommender system performs better in the aspect of similar neighbors grouping.The rating prompting mechanism is a combination of the product items rating process and a sales promotion mode——purchasing credit. Integrating the product item rating process into an entire e-commerce operation flow can well bestir the users to participate in the modified system by rating product items. Simultaneously, the services offered in the system will be more satisfactory.In the end, the application of and related solutions to the modified recommender system in my publishing enterprise is presented. The system's rationality and effectiveness are verified via open test on its performance. The thesis is a valuable reference for the publishing industry to boost e-commerce business with the modified personalized recommender system.