Sentiment Analysis on data breach topic tweets.
This project aims to conduct sentiment analysis of tweets related to terms "data breach" that hopefully will be useful for data breach detection.
We scraped the tweets and doing data cleansing (removing slang/punctuation, adding space, etc) in order to be able to get insight from it. Our datasets can be found in the datasets folder. We also used kamus to do the sentiment analysis.
We've done two approaches of vectorizer: using CountVectorizer and TF-IDF. Comparing those two performance with chosen classification algorithms (CountVectorizer using Random Forest, SVM and TF-IDF using Random Forest, SVM, and IndoBERT), we got the best result came from IndoBERT.
| Vectorizer Method | Model | Accuracy | F1 Score |
|---|---|---|---|
| CountVectorizer | Random Forest | 0.78 | 0.73 |
| CountVectorizer | SVM | 0.78 | 0.72 |
| TF-IDF | Random Forest | 0.74 | 0.67 |
| TF-IDF | SVM | 0.79 | 0.68 |
| TF-IDF | IndoBERT | 0.82 | 0.79 |
| File Name | Description |
|---|---|
| databocorv2_part2 | raw data from scraping |
| Scraping_preprocessing_v2 | output of notebook v1 |
| colloquial-indonesian-lexicon.csv | indonesian slang words dictionary |