Cluster Analysis of Restaurant Evaluation Features Based on UGC Data

With the advent of the big data era, analyzing using User Generated Content (UGC) data has become popular research. Most of the current research is to use UGC data to analyze user preferences or to analyze the influence of UGC data on other users. This paper, however, is based on the perspective of merchants, using UGC data to analyze the advantageous features of merchants that are recognized by users, and to help merchants improve their brand image. In this paper, the original UGC data is preprocessed through Chinese word segmentation, de-duplication and other techniques to generate a review UGC corpus, and then the Word to Vector (word2vec) model is used to train the UGC corpus to obtain the word vectors, and then the words are clustered into word clusters according to the word vectors, and a number of representative words are picked out from each word cluster to generate a thesaurus of restaurant evaluation features. In this paper, 34 restaurant evaluation keywords are finally extracted and labelled as a total of 14 categories of restaurant evaluation features.