project
NLP Project: Sentiment Analysis of Indonesia’s 2019 Presidential Election Tweets
As part of Team NLP A, I contributed to a project aimed at analyzing public sentiment on Twitter during the 2019 Indonesian presidential election. The goal was to classify 1,815 tweets into positive, neutral, or negative sentiments to gain insights into public opinion and sentiment trends during this major political event.
🎯 Project Objectives
- Analyze sentiment polarity (positive, neutral, negative) from Twitter data during Pilpres 2019.
- Compare multiple machine learning and deep learning models for sentiment classification.
- Derive insights into public opinion trends based on model predictions.
📂 Dataset
- Twitter dataset related to Indonesia’s 2019 Presidential Election (Pilpres).
- Preprocessed for noise reduction, tokenization, and normalization.
🚀 Models and Approaches
- TransformerEncoder
- IndoBERT (base-uncased and base-p2)
- RoBERTa (Indonesian version)
- LSTM and BiLSTM
- CNN-BiLSTM
- SVM Classifier
- Random Forest
- Gradient Boosting Classifier
📈 Evaluation Results (Accuracy)
| Model | Training | Testing |
|---|---|---|
| TransformerEncoder | 0.94 | 0.61 |
| IndoBERT (base-uncased) | 0.97 | 0.62 |
| IndoBERT (base-p2) | 0.66 | 0.42 |
| RoBERTa (Indonesian) | 0.72 | 0.59 |
| LSTM | 0.96 | 0.50 |
| BiLSTM | 0.99 | 0.54 |
| CNN-BiLSTM | 0.98 | 0.55 |
| SVM | 0.93 | 0.60 |
| Random Forest | 0.99 | 0.58 |
| Gradient Boosting | 0.82 | 0.56 |
🛠 Key Contributions
- Data preprocessing and exploratory data analysis (EDA).
- Model development and fine-tuning.
- Performance evaluation and comparative analysis.
👥 Team Members
- Aswin
- Wiliy
- Rohaldi
