Design And Implementation Of Text Classification Component In Quality Safety System

In recent years, quality safety incidents occurred in food, automobile and otherindustries frequently, the vital interests of consumers was suffered serious harm. With therapid development of Chinese economy and the complication of the supervision objects,the traditional method of sampling inspection is not enough to solve this growing problem.Internet is the main platform for information dissemination, the analysis of the massivequality safety information on the Internet, can contribute to the quality safety supervisionThis thesis mainly studies the design and implementation of text classificationcomponent in quality safety system, by using text classification technology to filter andanalysis massive feedback information from consumers. The research contents include:(1)Information Filtering, this part designed and implemented two binary classifiers, one isbased on the naive Bayes algorithm, the other is based on support vector machinealgorithm. The purpose is to filter the information which is unrelated to quality safety, andto ensure the accuracy of data analysis.(2) Automatic Classification, this part alsodesigned two multiple classifiers on naive Bayes and support vector machine algorithm.The multiple classifiers is used to classify the quality safety related information inautomatic industry, to specific the problem of quality safety, and to find out the potentialcrisis.(3) Test and analysis, The classifiers are evaluated by accuracy rate, recall rate, F1estimation and other assessments, to evaluate classification performance and analyze theeffect of application.The text classification component based on naive Bayesian and support vectormachine algorithm has been applied to the quality safety system. The result of testing andapplication showed that, the text classification component can extract quality safetyrelated information and find potential pitfalls, to achieve the purposes of quality safetysupervision and early warning.