Add Readme.md for description
Browse files
README.md
CHANGED
@@ -1,3 +1,137 @@
|
|
1 |
-
---
|
2 |
-
license:
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
- id
|
6 |
+
metrics:
|
7 |
+
- accuracy
|
8 |
+
pipeline_tag: text-classification
|
9 |
+
---
|
10 |
+
|
11 |
+
# Election Tweets Classification Model
|
12 |
+
|
13 |
+
This repository contains a fine-tuned of ***indolem/indobertweet-base-uncased model*** for classifying tweets related to election topics. The model has been trained to categorize tweets into eight distinct classes, providing valuable insights into public opinion and discourse during election periods.
|
14 |
+
|
15 |
+
## Classes
|
16 |
+
|
17 |
+
The model classifies tweets into the following categories:
|
18 |
+
1. **Politik** (2972 samples)
|
19 |
+
2. **Sosial Budaya** (425 samples)
|
20 |
+
3. **Ideologi** (343 samples)
|
21 |
+
4. **Pertahanan dan Keamanan** (331 samples)
|
22 |
+
5. **Ekonomi** (310 samples)
|
23 |
+
6. **Sumber Daya Alam** (157 samples)
|
24 |
+
7. **Demografi** (61 samples)
|
25 |
+
8. **Geografi** (20 samples)
|
26 |
+
|
27 |
+
## Libraries Used
|
28 |
+
|
29 |
+
The following libraries were used for data processing, model training, and evaluation:
|
30 |
+
|
31 |
+
- Data processing: `numpy`, `pandas`, `re`, `string`, `random`
|
32 |
+
- Visualization: `matplotlib.pyplot`, `seaborn`, `tqdm`, `plotly.graph_objs`, `plotly.express`, `plotly.figure_factory`
|
33 |
+
- Word cloud generation: `PIL`, `wordcloud`
|
34 |
+
- NLP: `nltk`, `nlp_id`, `Sastrawi`, `tweet-preprocessor`
|
35 |
+
- Machine Learning: `tensorflow`, `keras`, `sklearn`, `transformers`, `torch`
|
36 |
+
|
37 |
+
## Data Preparation
|
38 |
+
|
39 |
+
### Data Split
|
40 |
+
The dataset was split into training, validation, and test sets with the following proportions:
|
41 |
+
|
42 |
+
- **Training Set**: 85% (3925 samples)
|
43 |
+
- **Validation Set**: 10% (463 samples)
|
44 |
+
- **Test Set**: 5% (231 samples)
|
45 |
+
|
46 |
+
### Training Details
|
47 |
+
- **Epochs**: 3
|
48 |
+
- **Batch Size**: 32
|
49 |
+
|
50 |
+
### Training Results
|
51 |
+
|
52 |
+
| Epoch | Train Loss | Train Accuracy | Validation Loss | Validation Accuracy |
|
53 |
+
|-------|------------|----------------|-----------------|---------------------|
|
54 |
+
| 1 | 0.9382 | 0.7167 | 0.7518 | 0.7671 |
|
55 |
+
| 2 | 0.5741 | 0.8229 | 0.7081 | 0.7931 |
|
56 |
+
| 3 | 0.3541 | 0.8958 | 0.7473 | 0.7953 |
|
57 |
+
|
58 |
+
## Model Architecture
|
59 |
+
|
60 |
+
The model is built using the TensorFlow and Keras libraries and employs the following architecture:
|
61 |
+
|
62 |
+
- **Embedding Layer**: Converts input tokens into dense vectors of fixed size.
|
63 |
+
- **LSTM Layers**: Bidirectional LSTM layers capture dependencies in the text data.
|
64 |
+
- **Dense Layers**: Fully connected layers for classification.
|
65 |
+
- **Dropout Layers**: Prevent overfitting by randomly dropping units during training.
|
66 |
+
- **Batch Normalization**: Normalizes activations of the previous layer.
|
67 |
+
|
68 |
+
## Usage
|
69 |
+
|
70 |
+
### Installation
|
71 |
+
|
72 |
+
To use the model, ensure you have the required libraries installed. You can install them using pip:
|
73 |
+
|
74 |
+
```bash
|
75 |
+
pip install numpy pandas matplotlib seaborn plotly pillow wordcloud nltk tensorflow keras scikit-learn
|
76 |
+
```
|
77 |
+
|
78 |
+
### Data Cleaning
|
79 |
+
|
80 |
+
The data was cleaned using the following steps:
|
81 |
+
1. Converted text to lowercase.
|
82 |
+
2. Removed 'RT'.
|
83 |
+
3. Removed links.
|
84 |
+
4. Removed patterns like '[RE ...]'.
|
85 |
+
5. Removed patterns like '@ ... ='.
|
86 |
+
6. Removed non-ASCII characters (including emojis).
|
87 |
+
7. Removed punctuation (excluding '#').
|
88 |
+
8. Removed excessive whitespace.
|
89 |
+
|
90 |
+
### Sample Code
|
91 |
+
|
92 |
+
Here's a sample code snippet to load and use the model:
|
93 |
+
|
94 |
+
```python
|
95 |
+
import tensorflow as tf
|
96 |
+
from tensorflow.keras.models import load_model
|
97 |
+
import pandas as pd
|
98 |
+
|
99 |
+
# Load the trained model
|
100 |
+
model = load_model('path_to_your_model.h5')
|
101 |
+
|
102 |
+
# Preprocess new data
|
103 |
+
def preprocess_text(text):
|
104 |
+
# Include your text preprocessing steps here
|
105 |
+
pass
|
106 |
+
|
107 |
+
# Example usage
|
108 |
+
new_tweets = pd.Series(["Your new tweet text here"])
|
109 |
+
preprocessed_tweets = new_tweets.apply(preprocess_text)
|
110 |
+
# Tokenize and pad sequences as done during training
|
111 |
+
# ...
|
112 |
+
|
113 |
+
# Predict the class
|
114 |
+
predictions = model.predict(preprocessed_tweets)
|
115 |
+
predicted_classes = predictions.argmax(axis=-1)
|
116 |
+
```
|
117 |
+
|
118 |
+
## Evaluation
|
119 |
+
|
120 |
+
The model was evaluated using the following metrics:
|
121 |
+
- **Precision**: Measure of accuracy of the positive predictions.
|
122 |
+
- **Recall**: Measure of the ability to find all relevant instances.
|
123 |
+
- **F1 Score**: Harmonic mean of precision and recall.
|
124 |
+
- **Accuracy**: Overall accuracy of the model.
|
125 |
+
- **Balanced Accuracy**: Accuracy adjusted for class imbalance.
|
126 |
+
|
127 |
+
## Conclusion
|
128 |
+
|
129 |
+
This fine-tuned model provides a robust tool for classifying election-related tweets into distinct categories. It can be used to analyze public sentiment and trends during election periods, aiding in better understanding and decision-making.
|
130 |
+
|
131 |
+
## License
|
132 |
+
|
133 |
+
This project is licensed under the MIT License.
|
134 |
+
|
135 |
+
## Contact
|
136 |
+
|
137 |
+
For any questions or feedback, please contact [me] at [rendikarendi96@gmail.com].
|