Abstract: In the data-booming epoch, online reviews have become the scholars’ focus home and abroad due to its information diversity and its mass participation character. It aims at delving into the valuable consumption information contained in online comments, discriminating the product demand preference, and thus summarizing the customer segmentation and characteristics. Based on four selling smartphones on the Jingdong and Tmall online shopping platforms, 26489 effective online reviews are obtained as text data in this study. First, the features of mobile products with the decision algorithm of boundary average entropy (BAE) are extracted, and consumers’ product demand preference is classified and discriminated on the correlation analysis of mutual information and semantic similarity. Then, the scores are obtained on consumers preference discriminating the products’ seven dimensions according to the analysis of emotional tendency, meanwhile, a multidimensional score vector is formed to represent consumers. With the improvement of two-step cluster method being used, the classification of consumer groups and the summary of features are completed. Thereafter, the consumer groups of the four smartphones are analyzed and some related revelations are provided according to the research results. The research ideas and methods applied in this paper can be of vital reference and significance for enterprises to effectively discriminate the consumers’ product demand preference and scientifically classify the consumer groups.

Key words: online reviews; product demand preference; customer groups; decision algorithm of boundary average entropy; multidimensional score vector