Research And Application Of Recommender System In Cloud Contact Center

With the rapid growth of network information and the rapid growth of user needs,in order to give users a better service,more and more systems will introduce the recommendation services.The recommender system makes up the shortcomings of the search engine which needs the keyword query,constructs a good channel connecting the user and the service provider,on the one hand,solves the problem that the service provider information is displayed effectively,and on the other hand provides the user with the better personalized service.The recommender system has penetrated into the many Internet services we have come into contact with in our lives.But the recommender system is quickly and efficiently introduced into the existing system,for the vast majority of small and medium-sized service providers is still a problem.Cloud contact center as a platform to provide a variety of services,recommend the introduction of the system which is already inevitable.How to provide an open,customizable recommendation system in the cloud contact center that makes different applications improved is a very important issue.In view of the above mentioned problems,this paper designs and implements a recommender system in the cloud contact center.Service providers can use this recommender system real-time,dynamic deployment of its business,the recommended services access to existing systems,to provide users with recommended services.Due to the particularity of the different areas of the business,the service provider can experiment with the recommender system and optimize the recommendation process to improve the recommendation effect.In addition,the recommender system to meet the needs of different service providers,providing a variety of business scenarios,in the scene can be configured with different recommendation processes to achieve business,scenarios,processes,organic combination.The system also provides algorithmic customization services,in the standard context of the development of the required algorithm for the construction of the recommended process.In order to provide high-performance service capabilities,this article uses Spark as a distributed computing engine,Hive storage of bulk data,HBase provides online data storage.This paper first describes the development of the recommended system,and the use of related technology large data storage Hive,machine learning library Spark MLlib,timer clock Quartz introduced.And then elaborates on the design and implementation of the recommender system.The recommender system is divided into data collection and processing,off-line calculation,online computing,near-line correction,configuration of these five modules.Data collection and processing module to the collection of data conversion,inspection,merger and other operations,into the Hive.Offline calculation module for the calculation of large amounts of data,including the recommended calculation and effect calculation of two sub-modules.The recommended calculation is a core of the whole recommendation system.The design and implementation of the recommendation process,the control,the editing,the data transfer form between the nodes and the context of the offline algorithm are mainly designed and realized.The effect calculation provides a calculation of the effect indicators used to optimize and describe the recommended effect.Online calculation based on HBase real-time response capability,according to the results of off-line calculation,through the online computing process,real-time return to the recommended results.The near line correction provides two major features of the item information update and recommended results correction.The configuration module is the five main configuration sub-modules of the service provider's business,including business,scenario,process,algorithm and effect.The rich configuration information provides a high degree of customization for the system.Secondly,given the recommendation of the off-line calculation process in the home page,the algorithm is optimized.Using the back propagation neural network(bpnn)combined with svdpp method to achieve the recommended accuracy of the upgrade,and open source data set given test analysis,when recommended Top5,the accuracy can be increased by more than 10%.Finally,this paper uses intelligent routing in cloud contact center as the application,and tests the function and performance of the system to verify that the system meets the demand.