INFO 526 — Data Analysis and Visualization

Assignment 4: Cluster & PCA Analysis of Angry Birds Personality Data

Course

INFO 526 — Data Analysis and Visualization
Master’s in MIS/ML program, University of Arizona

---

Objective

To explore what defines a bird’s personality using cluster analysis and principal components analysis (PCA).
We used standardized physiological, morphological, and behavioral traits to test how well unsupervised clusters and PCA align with original Personality groups.

---

Methods

• Data Preparation: Standardized variables (Weight, Wing Length, Tarsus Length, cortisol measures, and behavioral traits).
• Cluster Analysis: Ran K-means (k=3) to group individuals into clusters.
• PCA: Extracted 4 principal components to identify the strongest axes of variation.
• Comparison: Plotted original Personality groups vs cluster assignments in PC1–PC2 and PC3–PC4 space.

---

Deliverables

1. Cluster Plot (4 panels):

   • Top row: Personality groups (PC1–PC2, PC3–PC4)
   • Bottom row: K-means clusters (PC1–PC2, PC3–PC4)
   • Caption highlights partial alignment and ARI ≈ –0.006 (random).

2. PCA Plot (4 panels):

   • Top row: Personality groups (PC1–PC2, PC3–PC4)
   • Bottom row: K-means clusters (PC1–PC2, PC3–PC4)
   • Caption highlights ~39% variance explained by PC1+PC2 and main trait drivers (cortisol, novelty, morphology).

3. Interpretation (answers):

   • Q1: Clusters do not match well with Personality groups.
   • Q2: Key separating traits are Cortisol (CORT1–3), Neophilia/Neophobia, body size traits (Weight, Wing, Tarsus), Exploration, Boldness, Aggression.
   • Q3: PCA aligns partially with Personality groups (especially along PC1), but not with clusters. Overlap remains across groups.

---

Key Results

• Clustering performance: Poor (ARI ≈ –0.006), showing little alignment with Personality groups.
• Most informative traits: Cortisol levels, novelty traits, size measures, exploration, boldness, and aggression.
• PCA insight: Captures meaningful physiological and behavioral axes of variation, but does not fully separate the three Personality categories.

---

Repo Structure

├── .gitignore
├── README.md
├── dataset
│   └── angry_birds_personalities.csv
├── docs
│   └── Week 5_Graded_Assessment_4.pdf
├── notebooks
│   ├── Cluster_Plot.png
│   ├── PCA_Plot.png
│   └── Week_5_Graded_Assessment_4.ipynb
└── requirements.txt

---