Evaluation of the Multinomial Algorithm in Predicting Technostress on Indonesian Fintech Adoption Sentiment

This paper evaluates the performance of a Multinomial Naïve Bayes algorithm for sentiment prediction using Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW) text transformation methods. The evaluation focuses on measuring the model's quality across different train-test data splits (80:20, 75:25, 70:30) to understand the model's effectiveness in predicting sentiment with varying training and testing data ratios. The dataset comprises user reviews from financial technology (fintech) services in Indonesia. The results highlight the differences in model performance across various data splits and text transformation methods. Additionally, this study explores technostress among Indonesian fintech users, providing insights into the unique challenges faced by this demographic. The evaluation is conducted using a confusion matrix to provide a comprehensive analysis of model accuracy and reliability.