论文部分内容阅读
【目的】针对电子商务平台的中文产品评论,提出一种面向供应链的客户关注特征挖掘方法。【方法】以产品评论数据预处理方法为核心,改进关联规则挖掘产品特征方法。预处理技术包括产品评价概念树、产品评价特征库和MA_Apriori算法。数据实验以京东商城平板电脑为例,在Weka环境中完成客户关注特征的挖掘。【结果】实验表明,对于相同的事务文件,采用数据预处理再进行关联规则的产品特征挖掘,特征查全率为90.5%,而关联规则挖掘方法查全率仅为71.4%。并且本方法可实现产品特征挖掘结果的层次化和规范化。【局限】需要进一步补充汉语分词系统的用户词典,添加产品领域相关的专业词汇,以提高分词准确性。【结论】本方法有助于供应链各节点企业灵活选择产品评价概念层次,从而有针对性地实施产品改进和服务提升。
【Objective】 Aiming at the Chinese product reviews of e-commerce platform, this paper proposes a customer-focused feature mining method for supply chain. 【Method】 Taking the method of product review data preprocessing as the core, this paper improves the association rules mining method of product characteristics. Pretreatment techniques include product evaluation concept tree, product evaluation feature library and MA_Apriori algorithm. Data Experiment Taking the Jingdong Mall tablet as an example, we completed the mining of the customer’s attention features in Weka environment. 【Result】 The experiment shows that for the same transactional documents, the product feature mining of association rules is performed using data preprocessing, the feature recall rate is 90.5%, while the correlation rule mining method is only 71.4%. And the method can realize the hierarchical and normalization of product feature mining results. [Limitations] User dictionaries need to be further supplemented with the Chinese word segmentation system to add related professional terms in the product field so as to improve the segmentation accuracy. 【Conclusion】 This method can help enterprises at all nodes in the supply chain to flexibly choose the concept hierarchy of product evaluation so as to implement product improvement and service improvement in a targeted manner.