<![CDATA[This study evaluates user sentiment towards the Bali Calendar application, analysing both positive feedback and negative critiques. The research employs the K-Nearest Neighbour (KNN) algorithm to classify sentiments as either positive or negative, aiming to assess overall public satisfaction with the app. To improve classification performance, the Tomek Links technique is applied in conjunction with KNN. The study categorizes data into pre- and post-COVID periods to address the observed increase in negative reviews following app updates during the pandemic. In the pre-COVID phase, KNN achieved accuracy rates of 93.7% and 94.3% with and without Tomek Links, respectively, using parameter values K=5 and K=3. In the post-COVID period, accuracy rates were 86.0% and 87.2% with and without Tomek Links, respectively, using parameter K=9. The application of Tomek Links resulted in a notable accuracy improvement of 1.2% in the post-COVID data compared to a 0.6% increase in the pre-COVID data. This finding highlights the significant role of Tomek Links in enhancing KNN accuracy, particularly when dealing with unbalanced datasets. The study demonstrates that while KNN performs robustly, Tomek Links can provide a substantial boost in classification accuracy, especially in scenarios with skewed data distributions.]]>
Copyrights © 2024
