电商平台用户评价文本数据统计分析方法研究

【作者】 刘阳；

【导师】 王楠；

【作者基本信息】 黑龙江大学 ， 应用统计， 2021， 硕士

【摘要】 本文以某电商平台销量情况为背景,以某平台商品文本评论数据作为数据支撑,分析电商平台用户评价文本数据对销量的影响,以及对不同类别商品用户的关注度情况。本文共包含251件商品信息,其中有效评论共77991条,在对文本评论信息处理后每条观测包含5个属性,分别为好评数量、中评数量、差评数量、未知评论类别数量以及平均商品评价字数。对不同类别的商品进行描述性分析,经分析发现电子类商品比较关注质量、外观、物流信息;非电子类商品比较关注味道、物流、包装信息。对背景数据进行统计分析方法研究,以产品销量作为因变量,好评数量、中评数量、差评数量、未知评论类别数量以及平均商品评价字数作为自变量,去探究影响产品销量的因素,故建立多元线性回归模型,所得到的R~2为0.54,建立随机森林回归模型,所得到的R~2为0.92,随机森林回归模型的结果相对多元线性回归模型的结果要好很多,得出结论为好评及差评数量对销量存在正向影响,中评及平均评价字数影响不如前两者。更多还原

【Abstract】 This article uses the sales volume of an e-commerce platform as the background,and uses the text comment data of a certain platform as the data support to analyze the impact of the e-commerce platform user’s evaluation text data on sales and the attention of users of different categories of goods.This article contains a total of 251 product information,including 77991 valid reviews.After processing the text review information,each observation contains5 attributes,which are the number of positive reviews,the number of moderate reviews,the number of negative reviews,the number of unknown review categories,and the average number of product evaluation words.A descriptive analysis of dif-ferent categories of goods is carried out.After analysis,it is found that electronic products pay more attention to information such as quality,appearance,and logis-tics;non-electronic products pay more attention to taste,logistics,and packaging information.Research on statistical analysis methods of background data,with product sales as the dependent variable,and the number of positive reviews,the number of moderate reviews,the number of negative reviews,the number of unknown review categories,and the average number of product evaluation words as independent variables to explore the main factors affecting product sales,So establish a multiple linear regression model.The obtained R~2is 0.54,Build a random forest regression model.The obtained R~2is 0.92.Random forest regression model is better than multiple linear regression model.From R~2and the regression coefficient,It is concluded that the number of positive reviews and negative reviews has a positive impact on sales,while the impact of middle reviews and average evaluation words is not as good as the previous two.更多还原