Clothing Product Reviews Mining Based on Machine Learning

Qing Hong, Peifei Feng, Zhichao Cheng

Abstract


This paper used the method of machine learning to study clothing product reviews classification based on big enterprise data. Taking Taobao clothing reviews as the object, it firstly excavated review themes from reviews corpus by association rules, and then searched review themes related to the categories by a method of mutual information to enrich the review themes. In the process of building classification models, commonly used SVM classifiers were studied in the beginning. After training and verification of a large amount of data, the classification accuracy reached 90.597%. In order to further improve the classification accuracy, the maximum entropy model was built by adopting the maximum entropy algorithm, on the basis of the same review themes. After repeated experiments and optimization in a large-scale of clothing product reviews, the classification accuracy reached up to 93.035% finally. Compared with SVM classification algorithm, the accuracy of maximum entropy in the clothing product reviews classification is higher. This paper verified the effectiveness of maximum entropy model on comment text multi-classification problem, and the maximum entropy model has practical values in electronic business.

Keywords


mutual information; review classification; SVM; the maximum entropy

Full Text:

PDF



International Journal of Online and Biomedical Engineering (iJOE) – eISSN: 2626-8493
Creative Commons License
Indexing:
Scopus logo Clarivate Analyatics ESCI logo IET Inspec logo DOAJ logo DBLP logo EBSCO logo Ulrich's logo Google Scholar logo MAS logo