Abstract:  Through data crawling, 1560 complaint texts in the comment area of China Mobile’s official microblog in sina are taken as ordinary complaint data and 4600 complaint texts about China Mobile from the black cat complaint platform are taken as further complaint data, and then the data is preprocessed. Word2vec technology is used to convert the preprocessed complaint text data into numerical data, and by calculating the vector cosine similarity, the ordinary complaints that are easy to become further complaints are obtained, and the themes are summarized. The LDA (linear discriminant analysis) subject analysis is used to obtain the subjects of the further complaints, so as to analyze their reasons. This study aims to provide new ideas for operators to accurately locate potential customers with further complaints, and put forward suggestions for operators’ customer complaint management.

Key words: further complaints, customer mining, text analysis