【摘要】 为应对日益缩短的产品设计周期要求,帮助某智能手表品牌厂家快速准确地收集用户反馈从而助力产品品质改进,对智能手表的网购评论进行数据挖掘。首先,从电商平台爬取该品牌智能手表用户评论,执行数据预处理后将评论文本转化为向量空间模型便于后续的情感倾向分析。为选用合适的情感倾向分析方案,研究选择了三种文本的向量化表示方案,结合四种机器学习分类算法对比分析了其在情感倾向分析方面的性能表现。实验结果证明了基于TF-IDF的文本向量化表示结合支持向量机做情感二分类方案的有效性,识别准确率在92%左右。更多还原

【Abstract】 This paper uses a data mining algorithm based on online shopping reviews to help a smart watch brand manufacturer quickly and accurately collect user feedback for improving product quality.It helps to cope with the increasingly shortened product design cycle requirements.Firstly,user reviews of the brand’s smartwatches are crawled from e-commerce websites.After performing data preprocessing,the comment text is transformed into a vector space model to facilitate subsequent sentiment analysis.In order to select an appropriate sentiment analysis program,this paper choses three vectorized representations of texts,combined with four machine learning classification algorithms to compare and analyze its performance in sentiment analysis.更多还原