02-09-2017, 01:22 PM
The collaborative allocation of logistics network resources can effectively meet the needs of customers. Maximize the overall benefit of the logistics network and ensure that the collaborative logistics network runs neatly at the time of value creation. Therefore, this article is based on the relationship of collaborative logistics network provider, transit warehouse and vendors, and we consider the uncertainty of time to establish a two-level programming model with random constraints and propose an intelligent algorithm simulated genetic annealing hybrid to solve it. A numerical example shows that the method has greater robustness and convergence; can achieve collaborative logistics network rationalization of resource allocation and optimization.
Recommendation systems are an important part of e-commerce, where appropriate articles are recommended to potential users. The algorithms most used to build recommendation systems in commercial applications are the collaborative filtering methods and their variants, which is mainly due to their simple implementation. In these methods, the structural characteristics of the bipartite network of users and elements are used and potential elements are recommended to users based on a measure of similarity that shows how similar the behavior of users is. In fact, the performance of memory-based CF algorithms depends to a great extent on the quality of similarities obtained between users / elements. As the similarities obtained are more reliable, a better performance is expected for the referral systems. In this paper we propose three models to extract the reliability of the similarities estimated in classic recommendations. We incorporate the reliability obtained to improve the performance of the referral systems. In the algorithms proposed for the extraction of reliability, several elements are taken into account, including the structure of the bipartite network of the user elements, the individual profile of the users, that is, how many items they have qualified and the elements, how many users have qualified them. Among the proposed methods, the method based on resource allocation provides the highest performance compared to others. Our numerical results in two sets of reference data (Movielens and Netflix) show that the use of resource allocation in classical recommendations significantly improves performance.
Recommendation systems are an important part of e-commerce, where appropriate articles are recommended to potential users. The algorithms most used to build recommendation systems in commercial applications are the collaborative filtering methods and their variants, which is mainly due to their simple implementation. In these methods, the structural characteristics of the bipartite network of users and elements are used and potential elements are recommended to users based on a measure of similarity that shows how similar the behavior of users is. In fact, the performance of memory-based CF algorithms depends to a great extent on the quality of similarities obtained between users / elements. As the similarities obtained are more reliable, a better performance is expected for the referral systems. In this paper we propose three models to extract the reliability of the similarities estimated in classic recommendations. We incorporate the reliability obtained to improve the performance of the referral systems. In the algorithms proposed for the extraction of reliability, several elements are taken into account, including the structure of the bipartite network of the user elements, the individual profile of the users, that is, how many items they have qualified and the elements, how many users have qualified them. Among the proposed methods, the method based on resource allocation provides the highest performance compared to others. Our numerical results in two sets of reference data (Movielens and Netflix) show that the use of resource allocation in classical recommendations significantly improves performance.