🏠 Min-Max Normalization & PCA Workshop

👤 Student Information

• Group: Group 6
• Team Members

• Emmanuel Ihejiamaizu (9080005)
• Liggia Elena Taboada Cruz (9085905)
• Chao-Chung Liu (9067679)

---

📘 Project Overview

This workshop explores two essential data preprocessing techniques in Machine Learning: Min-Max Normalization and Principal Component Analysis (PCA). We used a real-world housing dataset to understand how scaling affects data and how PCA helps find hidden structures in features.

🧠 Key Learning Topics

1. Min-Max Normalization

• Manual Implementation: We wrote the code from scratch without using libraries to understand the math: $\frac{x - x_{\min}}{x_{\max} - x_{\min}}$.
• Why it matters: Normalization ensures that features with large numbers (like Price) do not "bully" features with small numbers (like Bedrooms) during model training.

2. Principal Component Analysis (PCA)

• Standardization: We learned that StandardScaler is better than Min-Max for PCA because it gives every feature equal weight (Unit Variance).
• Variance Analysis: We found that 5 PCs are needed to explain 90% of the data. No single feature dominates this dataset.
• Loadings & Geometry: We used Heatmaps to see how features like Area_sqft and Lot_Size create a "tradeoff" in PC1.

🚀 Important Insights (Talking Points)

• Variance ≠ Prediction: Our most important finding was that high variance in PCA does not mean it can predict Price. The correlation was near zero (r ≈ 0.03).
• Non-Linearity: This suggests that house prices might have a non-linear relationship with the features, requiring models like Random Forest instead of simple linear PCA.

📂 Project Structure

.
├── data/
│   └── housing_data.csv        # The housing dataset
├── .gitignore                  # Git ignore file
├── MinMax_Normalization_Workshop.ipynb  # Main analysis notebook
├── README.md                   # Project documentation
└── requirements.txt            # Project dependencies.

🛠️ How to Run

1. 1. Clone the project:

   git clone [https://github.com/chooksemmanuel/Lab5---Min-Max-Normalization.git]

2. Create a Virtual Environment:

   python -m venv venv
   # On Windows:
   .\venv\Scripts\activate
   # On Mac/Linux:
   source venv/bin/activate

3. Install Dependencies:

   pip install -r requirements.txt

---