A novel data-driven weighted sentiment analysis based on information entropy for perceived satisfaction

Businesses have been devoted to improving customer satisfaction to avoid the user loss and sales volume drop, whereas the accuracy of relevant satisfaction research still remains a question. We aim to solve this question by addressing the four common issues: 1. Use consumers' perceived helpful information for satisfaction analysis (i.e. perceived satisfaction) to reduce the information overload and meanwhile effectively avoid the misleading arising from invalid information. 2. Present a distinctive prediction model to calculate perceived helpfulness to avoid the three biases caused by helpfulness voting method that is widely adopted in research of perceived satisfaction; 3. Take advantage of the uncertainty of information entropy to effectively avoid the problem that the satisfaction of new product features cannot be accurately mined and analyzed on account of frequency and quantity. 4. Calculate the perceived satisfaction results on weighted basis and conduct competition analysis in comparison with the results of congeneric products to further refine the satisfaction result. The findings of this study can help businesses to enhance the understanding of consumers’ satisfactions and preferences, and identify their dynamic market position with competitors of strategic planning for long-term development.