【作者】 丁勇；

【导师】 许炜；

【作者基本信息】 华中科技大学 ， 电子与通信工程， 2014， 硕士

【摘要】 近年来，我国频繁爆出重大的产品质量安全问题，这些问题不仅已经严重影响消费者的身心健康，甚至会影响社会安定和我国经济的长远发展。传统的质量安全监测手段主要是通过产品抽样检测、调查问卷、消费者投诉等方式完成，存在效率低下，覆盖面窄等缺点，难以应对这一日益严重的难题。随着互联网的发展，网络已经成为人们表达观点和发表言论的重要途径，这其中也包含消费者对产品质量的反馈信息，通过对这些反馈信息进行采集和分析，对质量安全问题的监测和预警有巨大的价值。本文设计并实现了一个分类系统，该系统利用文本分类技术对互联网上的质量安全信息进行自动处理，该系统主要功能包括：1）行业分类。采用支持向量机算法实现文本多分类器对质量安全信息按照所属行业进行分类，从而更好地对数据进行组织，便于对数据进行管理和查找。2）信息过滤。采用朴素贝叶斯算法实现文本二分类器对质量安全信息进行过滤，将无关的数据过滤，对数据进行提纯，为用户提供准确的质量安全数据。3）风险分类。采用支持向量机算法实现文本多分类器对各个行业的数据进行风险分类，将质量信息细化到具体的风险类别，便于对产品风险进行监测和预警。本文设计开发的文本分类系统经过测试分析表明，该系统能够对质量安全网络信息进行有效的处理，并对其他领域的文本分类系统有借鉴意义。更多还原

【Abstract】 In recent years, quality safety incidents occurred frequently in our country, it hasaffected not only people’s physical and mental health, but also the stability of society andeven the development of economy. The traditional supervisory systems of quality safetydepend on methods such as sampling inspection, questionnaire or customer complaints.These methods have a lot of disadvantages such as inefficiency and limited coverage, andthey are not enough to solve this growing problem. With the development of Internet, ithas become the main platform for people to express their opinions and comments, includea lot of comments of consumer on product quality safety. The collection and analysis ofthese data is valuable to the inspection and forecasting of quality safety.This thesis designed and implemented a text classifying system which can processthe massive online quality safety feedback information from consumer automatically byusing text classification technology. The system has the following functions:1) Industryclassifying. The system designed and implemented a text multiple classifier based onsupport vector machine algorithm to divide the data by industry and methodize the data, sothat the data is easy to manage.2) Information filtering. The system designed andimplemented a text binary classifier based on Naive Bayes algorithm to process the datafrom the crawler system. Its purpose is to filter the data which is not related to qualitysafety, clean the data and provide the accurate data to the users.3) Risk classifying. Thesystem also designed and implemented a text multiple classifier based on support vectormachine algorithm to specific the problem of quality safety and contribute to theinspection and forecasting of quality safety.The system which is designed and implemented by the thesis has passed the test now.The test and analysis show that the system can filter the unrelated data effectively andclassify data correctly and it has some reference for other classifying system.更多还原