KomDetek: Cyberbullying Detection in TikTok Comments

---

📖 Project Overview

KomDetek (Komentar Deteksi) is a Machine Learning and Natural Language Processing (NLP) project designed to tackle a growing issue in the digital age: online toxicity. Specifically focusing on the popular short-video platform TikTok, this project analyzes user comments to identify and classify cyberbullying and negative sentiment in the Indonesian language.

“Leveraging Machine Learning to foster safer and more positive digital communities on social media.”

🎯 Mission

To build a robust sentiment analysis classification model capable of understanding Indonesian internet slang, effectively distinguishing between normal commentary and harmful cyberbullying.

---

✨ Key Features & Methodology

🧠 Natural Language Processing (NLP)

• 🧹 Text Preprocessing - Advanced cleaning of raw TikTok comments (removing emojis, special characters, URLs, and transforming text to lowercase).
• 📖 Stopword Removal & Stemming - Utilizing NLP libraries to extract the root words and remove non-essential vocabulary.
• 🔠 Slang Normalization - Processing Indonesian internet slang into structured formats for better model understanding.

🤖 Machine Learning Pipeline

• 🔢 Feature Extraction - Transforming text data into numerical vectors using techniques like TF-IDF (Term Frequency-Inverse Document Frequency) or Count Vectorizer.
• 📈 Model Training - Training classification models to accurately categorize comments into specific sentiment classes.
• 📊 Performance Evaluation - Assessing model reliability using metrics like Accuracy, Precision, Recall, and F1-Score, complete with visual confusion matrices.

---

🛠️ Technology Stack

Category	Technology	Purpose
Language	Python 3	100% Core development language
Environment	Google Colab	Cloud-based Jupyter environment for ML
Data Manipulation	Pandas, NumPy	Data structuring and numerical computation
NLP Processing	NLTK, Sastrawi	Text processing for Bahasa Indonesia
Machine Learning	Scikit-Learn	Model building, training, and evaluation
Data Visualization	Matplotlib, Seaborn	Visualizing distributions and metrics

---

🏗️ The Machine Learning Architecture

NLP Data Pipeline

🛡️ KomDetek Workflow
├── 📥 1. Data Collection (Raw TikTok comments dataset)
├── 🧹 2. Text Preprocessing 
│   ├── Case Folding & Cleansing
│   ├── Tokenization
│   └── Stopword Removal & Stemming
├── 🧮 3. Feature Extraction (Text to Vector)
├── 🤖 4. Model Training (Classification Algorithms)
└── 📈 5. Evaluation (Confusion Matrix & Accuracy Scores)

---

🚀 How to Run the Project

Since this project is fully developed in Google Colaboratory, you can easily run and interact with the code without installing anything locally!

1. Open the Notebook: Click the Open in Colab button at the top of this page.
2. Copy to Drive: Go to File > Save a copy in Drive so you can make your own edits.
3. Run the Cells: Click Runtime > Run all to execute the data pipeline from preprocessing to model evaluation.

(Alternatively, you can clone the GitHub Repository and run the .ipynb file locally using Jupyter Notebook).

---

🌟 Future Development Plans

Phase	Feature	Timeline	Status
Phase 1	📊 Data Collection & Preprocessing Pipeline	Completed	✅ Finished
Phase 2	🤖 Machine Learning Model Training	Completed	✅ Finished
Phase 3	🕸️ Web Interface / Dashboard Deployment	Planned	📋 Planned
Phase 4	🧠 Deep Learning Integration (LSTM / BERT)	Planned	💭 Concept

---

📊 Project Highlights

Key Achievements

• ✅ Cloud-Native Development - Successfully utilized Google Colab for seamless model training and experimentation.
• ✅ High-Quality Preprocessing - Handled the complex, unstructured nature of social media text (TikTok comments) using robust normalization techniques.
• ✅ Accurate Classification - Built a model capable of distinguishing toxic behavior with high reliability.

---

💡 Conclusion

KomDetek proves that Machine Learning can be a powerful ally in the fight against online toxicity. By accurately detecting cyberbullying in TikTok comments, this project demonstrates a complete end-to-end Natural Language Processing pipeline.

By leveraging cloud environments like Google Colab, the barrier to reproducing this research is eliminated, allowing anyone to explore the model’s logic step-by-step. This project lays the groundwork for automated moderation tools that can help creators and platforms maintain a healthy, positive environment for their users.

---

🌐 Links & Resources

Developed with ❤️ by aipunpratama

  

---

_{🛡️ Supporting safer internet spaces. Give the repository a star if you support this cause! ⭐}