| As the development of information technology, the growth of the Internet and the popularity of network applications, they increase the amount of information received by users exponentially. The era of overload information is coming. In such an era, we receive the vast amounts of information from outside every day, but difficult for us to associate target information and filter them effectively. Whether to the information receiver orthe information providers, has encountered unprecedented challenges. The receiver how to find wanted information, and the provider how to put his information to right users,all whichproblems need to be solved together by the two sides. Traditional search engines are often stretched thin at this time, so the recommendation system is appeared. Recommendation system associates the user with the information by analyzing the user’s personal preference, andto help the user filtering and selecting information which might be useful to the user.eCRM is a decision support platform on the basis of big data, and which integrates CRM(Customer Relationship Management) with e-commerce(electronic Commerce) together.eCRM contains a large amount of data related to both products and users. It is the inevitable choice to design a recommend system based on eC RM, inwhich the enterprise product information and the user information are associated from each other. By this system,eCRM platform can give userreasonable and appropriate recommendation information according to the user’s interest, hobby or similar interests and buying behavior of other users, hence, it can avoid affliction on the time of selecting product from vast information.Recommendation service can not only greatly improve users’ satisfaction degree, increase customers’ loyalty and enterprise’ brand awareness, but also can fully embody the management philosophy of customer relations.Traditional recommendation system tends to run on a single machine, which greatly restricts the performance of system. O n the one hand, the time length needed by recommend result can’t meet the needs of users, and on the other hand, the increasing log data of users and computation loads are also restricted to the resources of single system.Aiming at the scalability problem, weintroduce fully parallelized idea to the recommend system. O n the basis of deeply research on Apache Hadoop’s Map Reduce and HDFS components, we deploy Hadoop on the cloud platforms before putting forward new parallelized and distributed network recommendation algorithms, and finally, a recommend systemprototype of eCRMis designed and implemented. In this paper, the main work done is as follows :1.A network recommendation algorithm and the Map Reduce process are designed on the basis of detailed study on the mechanism of Hadoop, theprinciple of parallel programming of MapReduce, common recommendation algorithms and systems, the substance diffusion algorithm and the heat conduction recommendation algorithm.2.Compliance with principles of high cohesion and low coupling, the modular architecture makes the recommendation system’s modules relatively independent.The recommendation system has better scalability by deploying it on the platform that taking Openstack as t he infrastructure service layer and Hadoop as the platform service layer.3. A high accuracy and novel recommendation system oriented to eCRM is designed, which is based on decision support platform of big data produced byeC RM, and using network recommendation algorithm designed by 1st work item. |
PDF Full Text Request