Predictive Modeling on Goodreads Dataset

Statistical Models for Business Analytics (Introduction to Machine Learning) - MGSC 310, Fall 2020

Corinne Smith, Erin Lee, Jon Le, Adam Gonzalez, Debbie Lu

Datasets

• books.csv
• final_dataset.csv

Note: Both Kaggle data sets were originally scraped from the Goodreads API.

Programming Language

R

Data At-A-Glance

• variables used in project:
  • outcome: “average_rating”
  • predictors (9 total):
    • "num_pages"
    • “book_ratings_count”
    • “text_reviews_count”
    • “title_sentiment_avg”
    • “authorworkcount”
    • “author_fans”
    • “author_ratings_count
    • “author_review_count”
    • “gender”

Note: removed 11 variables from original data set

Instructions

• Recommended: Run the program using the R Markdown format: "MGSC310FinalProject.Rmd" •

• Download the "datasets" folder.
• Download the "MGSC310FinalProject.Rmd" file.
• Given that the "datasets" folder & "MGSC310FinalProject.Rmd" file are in the same directory, open & run the R Markdown file in RStudio.

Models

1. Linear Regression

Table of Coefficients

Plot of Coefficients

2. Elastic Net

Plot of the Error Versus the Penalty (Regression with Regularization)

Plot of the Path of the Coefficients

3. Bootstrap Aggregated (Bagged) Decision Tree

Examining an Individual Model from Bagging

4. Random Forest

Plot of Error Vesus the Number of Trees

Variable Importance Plot

Plotting the Minimum Depth Distribution

Model Evaluation: Comparison of Metrics