YouTube Trending Project

Project Presentation Link

Analyzing data from the YouTube trending page in the US using data agreggated over October, December, and January in 2020 and 2021

Goal:

• To understand common characteristics of trending videos in the US

• Use this knowledge to build predictive models

• To predict engagement (likes and views) on a video in the US

Table of Contents:

• 1.Exploratory Data Analysis
  • 1.1 Data Analysis
    • 1.1.1 Raw Data Information
  • 1.2 Feature Engineering
    • 1.2.1 Like/Dislike Ratio
    • 1.2.2 Log of Ratings, Views, and Comment Counts
    • 1.2.3 Days Lapsed
    • 1.2.4 Duration Columns
    • 1.2.5 Title Length
    • 1.2.6 Tag Count
  • 1.3 Variable Analysis
  • 1.4 Heat Map
  • 1.5 Plotting Distributions
  • 1.6 Category Analysis
    • 1.6.1 Graphing video counts by category
  • 1.7 Correlation Matrix
• 2.Cleaning
  • 2.1 Preprocessing Data
    • 2.1.1 Drop Duplicate Rows
    • 2.1.2 Drop Columns
    • 2.1.3 Handling Missing Data
  • 2.2 Post-Processed Data
    • 2.2.1 Column Information
    • 2.2.2 Exporting Curated Data
  • 2.3 Export Cleaned Data w/ All Columns
• 3.Modeling
  • 3.1 Predicting Likes
    • 3.1.1 Pre-processing Data
      • 3.1.1.1 Train-Test Split (80:20)
      • 3.1.1.2 Initializing Pre-processing Pipeline
    • 3.1.2 Hyperparameter Tuning (Gridsearch)
    • 3.1.3 Regressors
      • 3.1.3.1 Linear Regression
      • 3.1.3.2 Random Forest
      • 3.1.3.3 XGBoost
    • 3.1.4 Random Forest
      • 3.1.4.1 Feature Importance
    • 3.1.5 Likes Evaluation
  • 3.2 Predicting Views
    • 3.2.1 Pre-processing Data
      • 3.2.1.1 Train-Test Split (80:20)
      • 3.2.1.2 Initializing Pre-processing Pipeline
    • 3.2.2 Hyperparameter Tuning (Gridsearch)
    • 3.2.3 Regressors
      • 3.2.3.1 Linear Regression
      • 3.2.3.2 Random Forest
      • 3.2.3.3 XGBoost
    • 3.2.4 Random Forest
      • 3.2.4.1 Feature Importance
    • 3.2.5 Views Evaluation