Data Mining | Dimensionality Reduction | Statistical Learning
This project explores how raw data becomes insight using the mathematical foundations of data mining and statistical learning.
Instead of relying on black-box machine learning libraries, this repository focuses on understanding, visualizing, and interpreting data using linear algebra and geometry.
Using the Iris dataset, this project implements and analyzes:
- Principal Component Analysis (PCA)
- Correspondence Analysis (CA)
- Multiple Correspondence Analysis (MCA)
- Distance-based data geometry
These are the same techniques used in feature extraction, clustering, and exploratory data analysis.
Most machine-learning projects show what a model predicts.
This project shows why the data behaves the way it does.
It answers questions like:
- Which variables matter most?
- How do classes separate geometrically?
- What hidden structure exists in the data?
This depth is what distinguishes a data scientist from someone who only runs libraries.
The project uses the Iris dataset containing:
- 150 samples
- 4 numerical features
- 3 flower species
It is one of the most widely used benchmark datasets in machine learning and statistics.
| Technique | Purpose |
|---|---|
| PCA | Reduces dimensionality while preserving variance |
| CA | Analyzes relationships between categorical variables |
| MCA | Extends CA to multiple categorical variables |
| Distance metrics | Measures similarity using Euclidean, Minkowski, and chi-square distances |
These form the mathematical core of modern data mining pipelines.
report/ -> Full project report with theory and experiments
results/ -> PCA, CA, and MCA visualizations
docs/ -> Methodology and explanations
The results folder contains:
- PCA projections of the Iris dataset
- Correspondence Analysis biplots
- Multiple Correspondence Analysis factor maps
These visuals show how meaningful structure emerges from raw data.
Sai Teja Bandaru
Data Scientist and Machine Learning Engineer
This project demonstrates an understanding of the mathematics behind machine learning, not just how to use libraries.
That skill is essential for data science, machine learning engineering, and research roles.