Artificial Intelligence Analysis Platform Design And Its Application

With the rapid development of big data technology,big data technology not only brings huge economic benefits to people,but also brings technical challenges to big data analysis and mining.The technology of the basic platform of big data is cumbersome,espe-cially the computing framework based on the X86 series processing platform has different characteristics and various interfaces,which greatly improves the technical difficulty for the developers of intelligent business applications based on the computing platform of big data.How to improve the efficiency of intelligent business application development,so that the application innovation based on big data platform can form reliable software and services according to the actual business needs,is a key issue in front of the software ser-vice providers of big data application innovation in all walks of life.In addition,repetitive data processing operations make the development cycle of big data business too long.This thesis studies the key technologies for the automatic generation of big data intelligent busi-ness,and designs and implements a visualized platform for the automatic generation of big data intelligent business.The specific work includes:(1)Aiming at the problem of complex types of business data in intelligent analysis of big data,this thesis designs a data format description method based on metadata manage-ment,and realizes the technical framework of standardized description and transmission of business logic metadata.Each subjob is defined as a complete data flow,and multiple component connector types are designed for flexible data flow processing.(2)To solve the problems of standardized development of big data business genera-tion platform and job code generation and construction,this thesis designs and implements visual job development and automatic code generation technology.A series of data pro-cessing operations are encapsulated as generic components,and visual business process development is carried out by dragging and dropping.In this thesis,two job templates,standard job and Spark job,are designed to meet the needs of users in different usage scenarios,and a code generation engine based on JET is designed for automatic code gen-eration of job flow.(3)This thesis verifies the job generation capability of the business generation plat-form through practical application cases.In view of the actual business requirements of telecom customer churn prediction and analysis,the visualization process is developed in the platform.This thesis designs and implements the component based on Cox propor-tional risk model,and completes a series of business processes such as data preprocessing,feature selection,model training and model evaluation.