Training attractive attribute classifiers based on opinion features extracted from review data

作者：

Highlights：

• The Kano model theory divides all the attributes of a product or service into 3 main categories: must-be, one-dimensional, attractive.

• We propose to use machine learning models to train predictive classifiers to automatically identify the attractive attributes of hotels from on-line review data.

• We propose a neural network model to extract features from review text to represent hotel attributes in the machine learning models.

• The experiment results show that the classifiers reach a precision of around 80% and outperform the existing statistical models by a margin of over 10%.

摘要

•The Kano model theory divides all the attributes of a product or service into 3 main categories: must-be, one-dimensional, attractive.•We propose to use machine learning models to train predictive classifiers to automatically identify the attractive attributes of hotels from on-line review data.•We propose a neural network model to extract features from review text to represent hotel attributes in the machine learning models.•The experiment results show that the classifiers reach a precision of around 80% and outperform the existing statistical models by a margin of over 10%.

论文关键词：Kano model,Attractive attributes,Review data,Neural networks,Machine learning

论文评审过程：Received 7 January 2018, Revised 27 August 2018, Accepted 8 October 2018, Available online 9 October 2018, Version of Record 22 October 2018.

论文官网地址：https://doi.org/10.1016/j.elerap.2018.10.003