Abstract: With the rapid development of e-commerce,the product review data is also rapidly expanding,which make it possible to obtain user attitudes from the review data and analyze the performance of product attributes.Effectively mining this information for design services is undoubtedly of great significance to product innovation.However,the existing product evaluation model lacks the consideration of the correlation between evaluation indicators,which easily leads to deviation of evaluation results.For this reason,a new method of improved interaction matrix was proposed.According to the frequency of attribute words mentioned in the review data,the user's value coefficient was determined for each attribute,and FPGrowth algorithm was used to mine the relationship between attributes.These two factors were used as the main diagonal and non-main diagonal of the matrix to form an improvement matrix,and a product evaluation analysis model was constructed based on this matrix.The relevance influence between the user's viewpoint and the evaluation elements was fully considered,and the deviation of the existing evaluation model structure was corrected.The comprehensive score of the product was calculated by analyzing the user's emotional attitude contained in the comment data.Taking the electric toothbrush on the B2C Website as an example,the feasibility and effectiveness of the proposed evaluation model was verified.

Key words: association relationship, interaction matrix, review data, product evaluation analysis model

摘要： 电子商务的高速发展伴随着产品评论数据快速膨胀,使通过评论数据来分析产品各属性表现成为可能,因此如何有效地挖掘这些信息用于产品设计具有重要意义。然而,现有的产品评价模型未考虑评价指标间关联关系,容易造成评价结果的偏差。为此,提出一种改进的相互作用关系矩阵的新方法。该方法依据评论数据中提及的属性词频率,确定用户对各属性的重视系数,利用FPGrowth算法挖掘属性间存在的关联关系,以这两方面因素分别作为矩阵主对角线与非主对角线元素形成改进矩阵,并基于该矩阵构建产品评价分析模型。该模型充分考虑用户观点与评价要素间的关联性的影响,修正现有评价模型结构的偏差。最后通过分析评论数据包含的用户情感态度,计算产品综合得分。以B2C网站上电动牙刷为例进行分析验证,证明了所构建评价模型的可行性及有效性。

关键词: 关联关系, 相互作用关系矩阵, 评论数据, 产品评价分析模型