PLS-SEM Approach and Sentiment Analysis for Identifying Significant Factors of Customer Satisfaction

Reviews have a direct impact on customer satisfaction. The aim of this study is to dissect and analyze a collection of 775 negative and 557 positive comment reviews, drawn from four distinct e-commerce platforms. By classifying these remarks into positive and negative sentiments, this research endeavors to illuminate underlying trends permeating these marketplaces. The research methodology employed involves field observations of online shopping experiences, utilizing data derived from 254 e-commerce customers. These data were collected via validated questionnaires and subsequently analyzed using the partial least squares structural equation modeling approach, employing the lavaan r library within the R programming environment. The questionnaire results produced a rating scale from 1 to 5, categorizing responses from 'very satisfied' to 'less satisfied', effectively illustrating both positive and negative commentary. The field comment data collected was coordinated with comment data extracted from four marketplace trading accounts. This data comment customer was analyzed using a range of comparative models such as k-nearest neighbors, multinomial naive bayes, stochastic gradient descent, and decision trees to conduct sentiment analysis. The findings reveal that the naive bayes method generates the greatest accuracy in sentiment analysis, registering an accuracy value of 0.886. Moreover, the analysis executed through r programming indicates that the e-service quality model yields the most robust results, reflected by an adjusted r-square value of 0.885. This study exerts a notable impact on service quality, as evidenced by a coefficient value of 0.865 and a perceived reputation score of 0.162.