YouTube-Analysis

A full-stack analytics solution designed to uncover patterns in content performance, audience engagement, and channel behavior across global YouTube trends. This project analyzes 100,000+ trending videos to empower content strategists, media analysts, and creators with predictive insights and interactive dashboards for smarter publishing decisions.

🚗 GitHub Project Repository

🔗 Click to view YouTube-Analysis

🧠 Project Overview

Understanding what makes a video trend is key to maximizing reach and engagement. This project delivers an end-to-end analytics platform that enables:

📊 Performance forecasting based on video metadata
👍 Engagement-driven content strategy
🌍 Country-specific trend analysis
📈 Dashboard-driven insights for creators and media teams

🎯 Key Objectives

Clean and preprocess multi-country YouTube data
Engineer features for view prediction and dashboarding
Build regression models to estimate video views
Deploy interactive dashboards for strategic content planning

📁 Project Structure

File Name	Description
`youtube.csv`	Raw dataset with trending video records
`cleaned_youtube.csv`	Preprocessed dataset with feature engineering
`youtube.sql`	SQL queries for data extraction and filtering
`sqlconnect.py`	Python script for SQL database connection
`youtube analysis.ipynb`	Jupyter notebook with EDA, modeling, and insights
`trending_model.pkl`	Trained regression model for view prediction
`trending_features.pkl`	Feature list used in modeling
`channel_encoder.pkl`	Label encoder for channel names
`country_encoder.pkl`	Label encoder for country names
`app.py`	Streamlit app for dashboard deployment
`YOUTUBE ANALYSIS DASHBOARD.accdb`	MS Access dashboard visualizing engagement and performance trends

🧹 Data Preprocessing

Encoded categorical variables (channel_title, country, category_id)
Extracted publish time features (hour, day, month)
Engineered engagement metrics (likes, dislikes, comment_count)
Removed duplicates and handled missing values
Normalized continuous variables for modeling

📈 Exploratory Data Analysis

📊 View distribution across countries and categories
👍 Engagement metrics over time
📺 Channel behavior and upload consistency
🧠 Correlation matrix of features influencing views
📅 Publish time impact on video performance

🤖 Modeling Approach

Target Variable: views
Algorithms Used: Linear Regression, Random Forest Regressor, XGBoost Regressor
Evaluation Metrics: MAE, RMSE, R² Score
Top Features: likes, comment_count, publish_hour, channel_title, country

📊 Dashboard Overview

🔷 MS Access Dashboard

Visualizes YouTube performance metrics and engagement trends:

🌍 View distribution by country and category
📈 Engagement metrics over time
📊 KPI cards for average views, likes, and comments

Power BI Preview

🟢 Streamlit App

Interactive dashboard for real-time video performance prediction:

📺 Input video metadata to forecast views
📊 Feature importance visualization
🌍 Country and category filters
📈 Engagement-driven prediction interface

Streamlit Preview

🚀 Deployment

Model serialized with joblib as trending_model.pkl
Dashboard deployed via Streamlit Cloud
SQL integration for dynamic video filtering
Git LFS used for large file management

🧠 Business Impact

Forecasts video performance for content planning
Identifies high-engagement formats and categories
Supports regional strategy with country-specific insights
Enhances creator decision-making with predictive analytics

🛠️ Tech Stack

Python: Pandas, NumPy, Scikit-learn, Streamlit
SQL: Data extraction and filtering
Visualization: MS Access, Matplotlib, Seaborn, Plotly
Deployment: Streamlit Cloud, GitHub, Git LFS

📌 Future Enhancements

Integrate NLP for sentiment analysis from video descriptions
Add clustering for channel behavior segmentation
Enable user-uploaded video metadata for prediction
Expand dashboard to include subscriber growth forecasting

👤 Author

Anesh Raj

🔗 GitHub Profile