Hijabistahub TikTok Sentiment Analysis — Engagement Insights & Model Benchmarking

This project analyzes TikTok comments related to Hijabistahub to understand public sentiment and engagement patterns, then benchmarks multiple text-classification models (Naive Bayes, SVM, Gradient Boosting). The objective is to convert unstructured social feedback into actionable insights that can support content strategy and brand perception monitoring.

---

What This Project Delivers

• Sentiment classification of TikTok comments (positive vs negative)
• Engagement insights using view-count trends and influencer comparisons
• End-to-end ML workflow (preprocessing → training/testing → evaluation → visualization)

---

Key Visuals (Project Overview)

Sentiment Distribution	Top Comment Terms
View Count Trend Over Time	Influencer vs View Count

---

RapidMiner Pipelines (Reproducible Workflow Evidence)

Training & Testing Workflow

Text Preprocessing Pipeline

---

Methodology (Summary)

1) Data Preparation

• Cleaned noisy social text (symbols, duplicates, inconsistent casing)
• Structured comments into a usable dataset for modeling

2) NLP Preprocessing

• Tokenization
• Case transformation
• Token-length filtering
• Stopword removal (English + additional filtering)

3) Model Benchmarking

The following models were trained and evaluated with consistent preprocessing:

• Naive Bayes
• SVM
• Gradient Boosting

---

Tools & Tech Stack

• RapidMiner Studio (workflow-based modeling and evaluation)
• Python (Jupyter Notebooks) for preprocessing / labeling support
• CSV datasets for training/testing inputs

---

How to Run (Fast Start)

RapidMiner

1. Open RapidMiner Studio
2. Load the .rmp processes
3. Ensure CSV paths are mapped correctly
4. Run the training/testing workflows and visualization process

Python

1. Open the notebooks (.ipynb)
2. Run cells in order to reproduce preprocessing and dataset preparation