<![CDATA[This study aims to evaluate the performance of a linear regression model in analyzing sentiment in text data in the form of tweets. The dataset used consists of tweets that have undergone text preprocessing, such as removing URLs, mentions, symbols, and numbers, as well as stemming and tokenization. The purpose of this preprocessing is to improve the quality of the feature representation in the form of TF-IDF, which is used as model input. The evaluation was conducted by comparing the model's performance on raw and cleaned data. The evaluation results show that the linear regression model has a Mean Squared Error (MSE) of 0.1597 and an R² Score of -1.2884, indicating that the model is unable to effectively explain data variability. Visualization of the comparison between predicted and actual scores reinforces this finding, indicating that the model struggles to capture the nuances of informal language, irony, and emotional context in tweets. In conclusion, linear regression is not an ideal approach for text-based sentiment analysis, and the use of contextual representation methods such as word embedding or BERT, along with non-linear predictive models, is recommended for more accurate and relevant results.]]>
Copyrights © 2025
