Research on personalized user-centered product improvement based on sentiment mining of online reviews and competitor analysis

Abstract

Online reviews contain a wealth of information about product performance and user sentiment, which can effectively assist manufacturers to make production decisions. Taking abundant online reviews as data, this study mined product attributes and consumer sentiments to analyze users' personalized preferences and the performance of competitive products, and proposed personalized user-centered product improvement strategies for differentiated markets. Specifically, the word2vec technology and the LSTM neural network were applied to develop the sentiment mining method that was suitable for specific product domain, through which the explicit emotional information was extracted from product reviews and the missing emotional information was also predicted combined with the overall sentiment of the review text and the attitude of other consumers. On this basis, the k-means method was applied in this model firstly to segment the users according to their emphasis on product attributes. Then, a preference identification model that describes the relative importance of product attributes was designed according to the utility function and KANO theory, so as to measure consumers’ various preferences of each user group. Finally, combined with the relative performance of product attributes, the traditional IPA model was optimized to determine the improvement priority of attributes for different user groups to meet user preferences in a differentiated way. These work helps enrich the relevant theories and methods of sentiment mining and provides practical guidance for user-centered product improvement.