摘要
产品属性权重表示消费者对属性的重视程度,是消费者选择产品和商家进行产品优化的依据,但目前基于在线评论确定属性权重的方法存在主观性较大和结果易偏差的弊端,因此文章提出了一种融合模型,首先基于改进LDA主题模型挖掘出产品属性,然后基于属性情感分析对评论中属性满意度进行标注。之后先对满意度标注样本进行N折交叉处理,然后计算所有不同样本下各属性的信息增益值,以此对各属性重要性值分布进行参数估计。然后,计算两两分布的距离因子,最后运用改进AHP方法确定属性权重。以肉类生鲜产品为对象进行实证分析,结果表明,所提出的融合模型能够准确计算出产品各属性权重。
Product attribute weight indicates the importance consumers attach to attributes, and is the basis for consumers to select products and businesses to optimize products. However, the current method of determining attribute weight based on online reviews has the disadvantages of subjectivity and easy deviation of results. Therefore, this paper proposes a fusion model. First, the product attributes are mined based on the improved LDA topic model, and then the attribute satisfaction in the reviews is marked based on attribute emotion analysis. After that, the satisfaction labeling samples are processed with N-fold crossover, and then the information gain value of each attribute under all different samples is calculated, so as to estimate the distribution of the importance value of each attribute. Then, the distance factor of pairwise distribution is calculated, and the attribute weight is determined by using the improved AHP method. The empirical analysis of fresh meat products shows that the proposed fusion model can accurately calculate the weight of each product attribute.
出处
《运筹与模糊学》
2023年第2期1176-1185,共10页
Operations Research and Fuzziology
