Sentiment Analysis and Satisfaction Evaluation of Smart and Connected Products Considering Interpretability
DU Yinfeng
ZHANG Jiamin
WANG Wei
Abstract：Smart and connected products(SCPs)integrate Internet of Things(IoT)and artificial intelligence technologies,relying on embedded sensors and network connections to achieve real-time interaction,and have become a core driving force for the development of smart homes,Industry 4.0 and smart cities.As an important data source that reflects consumer needs and satisfaction,user reviews provide essential support for product design optimization and user experience improvement.To address the limitations of existing sentiment analysis methods in complex semantic modeling,fuzzy linguistic quantification and interpretability,this paper proposes a comprehensive sentiment analysis framework and satisfaction evaluation method.Firstly,a bidirectional long short-term memory(BiLSTM)network is used to accurately capture semantic dependencies of review texts.Secondly,the probabilistic linguistic term set(PLTS)is utilized to quantify the uncertainty of fuzzy expressions.Furthermore,the contribution of features to sentiment prediction is revealed through Shapley additive explanations(SHAP).Finally,SHAP is combined with importance-performance analysis(IPA)to generate optimized priority recommendations.Taking smart speakers as an example,the effectiveness of the proposed sentiment analysis framework is verified by experiments.It achieves an F1 score of 0.967 and an AUC of 0.979 on the test set,significantly outperforming traditional methods.Based on the results,this paper not only helps to improve the sentiment analysis of SCPs,but also provides scientific basis and practical guidance for product function optimization and personalized recommendations.
Keywords：smart and connected products(SCPs)sentiment analysisimproved importance-performance analysis(IPA)bidirectional long short-term memory(BiLSTM)networkprobabilistic linguistic term set(PLTS)Shapley additive explanations(SHAP)
Publication Date：2025-12-31
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：11( 66-76 )