Improving the functional performances for product family by mining online reviews

Companies continuously perfect product performances directing at consumers’ feedback, seeking to enhance customer satisfaction and product competitiveness. To make up for the insufficiency of previous research on product family performance improvement, a method applies multiple data-mining techniques to dig out online reviews is put forward to quantify the improvement priority of each performance in the product family, so as to guide product family improvement. Web Crawler is employed to collect customer reviews of various product variants, and then natural language processing technology is utilized to identify the words expressing functional performances and customer sentiments in the reviews, where the term frequency of each performance is defined as its importance factor. The mapping model between performance specifications and module instances in the product family is established to obtain the commonality factor of each performance. Lexicon-based machine learning is exploited to analyze customers’ sentimental values for each performance specification, which is regarded as satisfaction factor. According to the importance and satisfaction of each performance, Kano coefficient is assigned to each performance by utilizing the Kano model. Finally, combined the three factors and Kano coefficient, the improvement priority of each performance specification is estimated to suggest the enterprise to plan the resource allocation for product family improvement. The feasibility of the proposed method is demonstrated by performance improvement for sweeping robot product family and comparison with traditional questionnaire method.