| With the rapid development of society,machine learning has been applied to various fields of computer science,and more and more companies and institutions are vigorously developing machine learning technology to empower enterprises to become intelligent.However,machine learning problems belong to a more professional field.A series of problems such as model building ? parameter tuning ? and model deployment make companies invest more in human development costs,therefore,there is an urgent need for an one-stop intelligent application development platform to assist in development Personnel work,so that ordinary developers can get rid of the cumbersome and engineered development model,and focus their energy on the iteration of algorithms and strategies.Based on the above background,the topic of this article comes from Enterprise commissioned research and development projects,an intelligent application development platform was designed and implemented.The platform is an one-stop intelligent application service platform for AI engineers.The platform integrates more than 80 algorithm components,and supports user-defined algorithm components,.Users only need to drag and drop related algorithm components to draw a flowchart to perform visual modeling,and it can be intuitively check the training results to identify the quality of the model and formulate further optimization plans.The platform provides users with support for the entire process operation of preprocessing data ?building models?training models?evaluating models?deploying model services and model invocation to support intelligent application development.It supports multiple algorithm frameworks to meet the needs of multiple AI application scenarios.The main works of the thesis are as follows:(1)Demand analysis.According to demand analysis,the intelligentapplication development platform is divided into workflow engine module?process scheduling module?process management module?model upload and API export module?data import module?online IDE module.Use the data flow diagram to describe the interaction process between the sub-modules in the system and between the user and the system.Use UML use case diagrams to describe process management?process scheduling?model import and call?data import and other functions to be implemented by the system,and describe the system's non-functional requirements according to actual needs.(2)Design and implementation.Combined with platform requirements,design the overall architecture of the intelligent application development platform and decompose the functions of the system,design and improve the relevant source code of the Activiti workflow engine framework,and implement operations such as suspending?activating?and running parallel gateways;implement the management and scheduling of the saved processes in the platform;implement data import and model management related work;implement using container technology to provide users with online programming services.In the entire design and implementation work,class diagrams and sequence diagrams are used to describe the implementation process of each function in detail.According to the characteristics of the data,different types of databases are used to store the platform data,HBase is used to store training result data,and Redis is used to store deep learning iteration data,My SQL stores business data,and performs the logical design and physical design of the database.(3)System testing.After completing the coding work,design test cases for the system,and complete the functional test of the system according to the test cases.At the same time,test for non-functional requirements such as system stability and response time.After system testing,the platform's various functions can work normally,the user interface can be used normally,and the business functions meet the requirements.The platform has sufficient functional correctness and robustness to meet the needs of an enterprise for an intelligent application development platform. |
PDF Full Text Request