Research And Implementation Of An Integrated Simulation Platform To Perform The Recommended System Container

Development of Internet technology has greatly promoted the way people interact with the network frequency, causing the amount of data in the network interactions exponentially, large amount of data generated also cause this serious problem of information overload. Many tools have been developed to assist in retrieving, searching, and filtering to solve the information overload problem, but these tools are more likely to provide much worthless information, and recommender systems is to overcome this limitation. Recommender systems filters in many resource materials or information users might be interested in, but there are limitations, existing recommender systems generally contain only a recommendation algorithm for a class, play a good role under certain conditions, but it does not apply to other conditions. This thesis presents a design scheme to implement a recommender systems simulation platform can simulate various types of algorithms freedom spots, a variety of algorithms to achieve coexistence mechanisms, and provide various types of algorithms corresponding to various types of data sets, and finally be able to use the environment is self-adaptive automatically detects and selects the appropriate algorithm.This thesis aims at the recommender systems integrated simulation platform execution engine adapted for a variety of algorithms, as well as its scalability for design analysis and implementation. First, the status of the development of the recommender systems are summarized, and the construction of related technical recommender systems simulation platform research and analysis, research parallel computing, cluster management and distributed framework is currently more popular. Then the actual demand for the simulation platform, focusing on the implementation of a detailed needs analysis engine; according to the overall design of the overall structure of demand, the decomposition algorithm execution and implementation of the container body configuration tools, namely the detailed design of its key features, and finally realized the coexistence of a variety of recommendation algorithm and can freely swap recommended simulation platform.The advancement of this thesis proposed the recommender systems simulation platform is mainly reflected in the following aspects:(1) The implementation of container-based distributed simulation platform is built framework Fourinone make full use of the existing framework Fourinone, and add the corresponding functions according to the actual needs of the simulation platform, with a personalized functional design; the use of distributed parallel computing process execution computational efficiency compared to stand-alone operation has significantly improved.(2) The simulation platform execution engine algorithms library default to store multiple types of algorithm, and the algorithm can be freely swap based on user needs, to achieve a variety of recommendation algorithm coexistence mechanisms, and provide different data sets corresponding algorithms for calls, and the algorithm can implementation of quality evaluation.(3) The simulation platform for the implementation of the task server using cluster management system and a variety of monitoring mechanisms used in a heartbeat cluster management control mechanisms other servers in the cluster tracking and monitoring the implementation of task assignments server server for execution progress monitoring, feedback to the user to perform the progress of the case.