The Design And Implementation Of User Feedback DC Based On Machine Learning

The project is derived from the actual project of Baidu Inc Dumi product line,belonging to the field of artificial intelligence in the internet.Dumi is an outstanding representative of the new generation of intelligent operating system,based on NLP technology,clear user needs,provide the corresponding services.In the product line,every day to receive one hundred thousand orders of magnitude user comments and user feedback,the amount of data is very large.Through classifying and screening of user feedback,users can get the issues related to the current product experience and suggestions,which reflects the existing problems and optimized version of the product parts,so as to guide the iterative requirements,but also provide the basis for quality assurance personnel tracking line.A large number of user feedback data text classification and screening is the key to the problem,but the current is to manually export some of the data from the online database,and artificial classification screening useful feedback.The use of machine learning methods,design and implementation of user feedback data center platform,the user feedback data into the platform,can efficiently and accurately on the huge volume of data and user feedback text classification screening,classification and statistics show,convenient to check and follow up the relevant person and the reasons for the investigation of user feedback the solution.User feedback data center platform system can be divided into three parts,the user feedback data pull,machine learning based on the classification of feedback data filtering and user feedback data center based on PHP and MySQL.Among them,the user feedback data pull Python prepared by relevant polling API from the company unified user feedback all the feedback data platform pull the product line and according to the need to re organize the data format and stored in the Hbase;feedback data classification using machine learning algorithm selection in learning,extract feature words optimization,classification and data according to the feature data classification corresponding screening;Data Center for data classification show,query terms,feedback tracking function.In this paper,the user feedback data center platform system needs analysis,overall design,detailed design,testing and verification work.I participated in the design and development of user feedback data pull,machine learning based on the classification of feedback data filtering and data platform related functions.The user feedback platform of data center has been put into use on the line at present,data classification rate reached 85%,the data center has greatly improved the efficiency of user feedback processing,and the release of the human data obtained,department leaders and colleagues alike.