Classifier-Chain SVM for TAM-Driven Multilabel Sentiment Analysis in Large-Scale User Acceptance Evaluation

Abstrak

In this work we suggest an alternative and novel integrated method which combines TAM with multilabel sentiment analysis by a SVM classifier. In contrast to literature on traditional TAM research that often comprises survey-based measures, this study utilises user-generated comments written in large numbers as a close substitute factor for perceived usefulness, perceived ease of use, attitude and behavioural intention. A set of 22,366 reviews were harvested from the Google Play Store for the video-editing application CapCut and handled within the Knowledge Discovery in Databases (KDD) framework that comprised data selection, preprocessing, labeling, transformation, mining and evaluation. It was observed that the multilabel SVM model trains 80% of Corpus19 and tested on remaining 20% achieves F1-score: 0.95 (Train) and 0.91 (Test). Stronger associations between both Perceived Usefulness and Perceived Ease of Use, and Attitude Toward Using (r = 0.912, 0.816 respectively) were again shown by Pearson correlation analysis. Meanwhile, Attitude Toward Using was positively related to Behavioral Intention of Use (r = 0.510). These results suggest that the TAM dimensions of comments can be efficiently modeled as multilabel sentiments, and usefulness and ease of use continue to play a significant role in shaping user acceptance towards CapCut

Identifier

• OpenAlex: https://openalex.org/W7169882099
• DOI: https://doi.org/10.67561/jaids.v1i1.3