CSC466 Final Project

Summer Mariana Ortega (sorteg16@calpoly.edu)
Diego Melgoza (drmelgoz@calpoly.edu)

Overview

This project investigates the logic behind dog breed popularity in the United States using American Kennel Club (AKC) breed traits, breed groups, and popularity rankings from 2013 to 2025.

Research Questions

1. Can dog breed traits predict a breed's popularity tier in the United States?

   • Model approach: Random Forest classification
   • Target variable: Popularity Tier

2. Can dog breeds be grouped into meaningful clusters based on their traits? Do those trait-based clusters align with the official AKC breed groups?

   • Model approach: Clustering
   • Comparison label: AKC Group

Data Files

• data/raw/raw_dog_traits_table_2013_2025.csv: combined AKC source data before project-specific cleaning.
• data/interim/breed_traits.csv: breed trait table.
• data/interim/breed_ranks.csv: breed popularity rankings from 2013 to 2025.
• data/interim/breed_groups.csv: AKC breed group labels.
• data/processed/breed_forest_full.csv: Question 1 modeling data with all individual trait ratings.
• data/processed/breed_forest_avg.csv: Question 1 modeling data with AKC trait-group average ratings.
• data/processed/breed_clustering_full.csv: Question 2 clustering data with all individual trait ratings plus comparison labels.
• data/processed/breed_clustering_avg.csv: Question 2 clustering data with AKC trait-group average ratings plus comparison labels.

Repository Structure

• data/raw/: original combined source data.
• data/interim/: cleaned intermediate data tables used to build modeling datasets.
• data/processed/: final modeling datasets only.
• models/: models from CSC 466 labs.
• notebooks/: project notebooks for cleaning, setup, modeling, and analysis.
• results/: model outputs.

Question 1 Modeling Data

For data/processed/breed_forest_full.csv and data/processed/breed_forest_avg.csv

• -1: row identifier, ignored by the classifier (Breed).
• 0: numeric predictor.
• positive integer: categorical predictor or class variable with that many possible values.

The class variable is Popularity Tier, currently defined as three average-rank groups

Question 1 Model Files

• notebooks/random_forest_full.ipynb: evaluates the custom Random Forest model using all individual trait ratings.
• notebooks/random_forest_avg.ipynb: evaluates the custom Random Forest model using trait-group average ratings.

Question 2 Modeling Data

For data/processed/breed_clustering_full.csv and data/processed/breed_clustering_avg.csv

• 1: numeric column used by the clustering algorithm.
• 0: metadata or ground-truth label column not used as a clustering feature.

The comparison label is AKC Group

Question 2 Model Files

• notebooks/clustering_full.ipynb: runs k-means clustering using all individual trait ratings and compares clusters to AKC groups.
• notebooks/clustering_avg.ipynb: runs k-means clustering using AKC trait-group average ratings and compares clusters to AKC groups.