FISAToon: Sentiment Analysis and Popularity Prediction for Webtoons

Project Duration: 2025.03.11 - 2025.03.13

This project analyzes the sentiment of webtoon comments and uses the results to predict webtoon popularity trends.
It explores the correlation between emotional shifts in user comments and changes in webtoon ratings.

• Tech Stack: FastAPI, MongoDB, Selenium, HTML, CSS, JavaScript
• Team Members (3 total): Fullstack (me), AI 1 (Sentiment Model), AI 2 (Prediction Model & Visualization)
• GitHub: FISAToon

📌 Project Overview
- Crawled best comments from Naver Webtoons to conduct sentiment analysis (positive/negative trends)
- Built machine learning and deep learning models to predict ratings based on sentiment and compare them to actual ratings
- Visualized insights through a web-based dashboard

🙋‍♂️ My Role & Contributions
- Developed automated crawling and preprocessing pipeline for webtoon comment data
- Built FastAPI backend and designed frontend web pages
- Designed MongoDB schema for unstructured data
- Implemented Selenium-based automation with remote debugging mode
- Handled full frontend (HTML, CSS, JS) and structured the overall service architecture

📘 Key Features
- Webtoon Sentiment Analysis: Used Hugging Face's "korean_sentiment" model for classification
- Popularity Prediction Models: Applied Random Forest, Linear Regression, and LSTM for prediction
- Visualized the following:
  ① Sentiment trends per webtoon
  ② Average ratings by genre
  ③ Average ratings by day of the week
  ④ Model performance comparisons (actual vs. predicted ratings)

📊 Data Visualization Highlights
- Sentiment chart: Visualizes positive/negative comment trends per webtoon
- Genre-based rating chart: Compares average ratings across genres
- Weekly rating chart: Analyzes rating trends by weekday
- Prediction performance: 3D scatter plot comparing actual and predicted ratings from different models