You can deploy this model and start earning money today!

feverlash/
Indonesian-SentimentAnalysis-Model

Name: Indonesian-SentimentAnalysis-Model
Rating: 2.5 (15 reviews)

bert

text-classification

kornwtp/indonlu-smsa

indobenchmark/indobert-base-p1

finetune

Views

No views yet

Indonesian Text Sentiment Analysis 🚀

📌 Overview

This project fine-tunes a transformer-based model to analyze sentiment for Indonesian text.

📥 Data Collection

The dataset used for fine-tuning was sourced from IndoNLU Datasets, specifically:
SmSA (IndoNLU) Dataset

🔄 Data Preparation

Tokenization:
- Used Indobert for efficient text processing.
Train-Test Split:
- The Dataset is already splitted into train, validation, and test.

🏋️ Fine-Tuning & Results

The model was fine-tuned using TensorFlow Hugging Face Transformers.

📊 Evaluation Metrics

Epoch	Train Loss	Train Accuracy	Eval Loss	Eval Accuracy	Training Time	Validation Time
1	`0.2471`	`88.15%`	`0.2107`	`91.31%`	`7:55 min`	`10 sec`
2	`0.1844`	`90.41%`	`0.2107`	`92.39%`	`7:50 min`	`10 sec`
3	`0.1502`	`91.66%`	`0.2135`	`93.14%`	`7:51 min`	`9 sec`
4	`0.1285`	`92.50%`	`0.2192`	`93.69%`	`7:50 min`	`10 sec`
5	`0.1101`	`93.13%`	`0.2367`	`94.14%`	`7:48 min`	`9 sec`

⚙️ Training Parameters

epochs = 5

learning_rate = 5e-5

seed_val = 42

max_length = 128

batch_size = 32

eval_batch_size = 32

🤖 How to use

python

import tensorflow as tf
from transformers import TFAutoModelForSequenceClassification, AutoTokenizer

# Load model dan tokenizer
model_name = "feverlash/Indonesian-SentimentAnalysis-Model"  # Ganti dengan path model yang telah disimpan
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = TFAutoModelForSequenceClassification.from_pretrained(model_name)

# Fungsi untuk melakukan prediksi sentimen
def predict(text):
    sentiment_mapping = {
        1: "positive",
        0: "negative",
        2: "neutral"
    }
    
    # Tokenisasi teks
    inputs = tokenizer(
        text,
        return_tensors="tf",
        truncation=True,
        padding="max_length",
        max_length=128
    )
    
    # Prediksi menggunakan model
    outputs = model(inputs)
    logits = outputs.logits

    # Menghitung probabilitas
    probabilities = tf.nn.softmax(logits).numpy()
    
    # Menentukan label prediksi
    predicted_index = int(tf.argmax(probabilities, axis=1).numpy()[0])
    predicted_label = sentiment_mapping.get(predicted_index, "unknown")
    
    # Keyakinan prediksi
    confidence = probabilities[0][predicted_index]

    print(f"Teks: {text}")
    print(f"Prediksi label: {predicted_label} (Confidence: {confidence:.2f})")

# Contoh penggunaan
text = "aku sedang jalan-jalan di Yogyakarta"
predict(text)