Enhancing accuracy of innovative topic modeling of cinema reviews in Twitter data through Latent Dirichlet Allocation in comparing non-negative matrix factorization
Krishna K. Leela, R Surendran · 2025
To enhance the accuracy and precision of movie reviews on Twitter by employing the Latent Dirichlet Allocation model and comparing its effectiveness with the Non-Negative Matrix Factorization technique. There are two groups of samples in the research study. The first group Latent Dirichlet Allocation includes 21 sample inputs and the group Non-Negative Matrix Factorization includes 21 samples, alpha value of 0.05 beta value of 0.2 and 80% G power is also being taken into consideration and the total sample size is 42. The accuracy of the Latent Dirichlet Allocation is 90% typically higher compared to 86% accuracy of the Non-Negative Matrix Factorization. The mean accuracy of the detection was found to be within +/- 2 standard Deviation, and a t-test using an independent sample showed that this result is statistically significant and got a significance value of p 0.009 (p<0.05). Performance of Latent Dirichlet Allocation is better than Non-Negative Matrix Factorization by comparing their results. The accuracy of Latent Dirichlet Allocation is higher than the comparison algorithm Non-Negative Matrix Factorization accuracy.