Object-Oriented Analysis of Scikit-Learn Library

This project presents an Object-Oriented Programming (OOP) analysis of the Scikit-learn library. The main purpose of this project is to understand how Scikit-learn uses important OOP concepts such as inheritance, polymorphism, abstraction, and composition in its internal architecture. This repository also includes a custom machine learning extension built using standard Scikit-learn interfaces.

• Muhammad Bilal — F25BDATS1M02098
• Maryam Tahir — F25BDATS1M02051
• Jaweria Zafar — F25BDATS1M02090

Submitted to: Dr. Akmal Shahbaz
Department: Department of Data Science, The Islamia University of Bahawalpur
Subject: Object-Oriented Programming (BS Data Science — 2nd Semester — M2 Section)
Session: 2026

This project bridges the gap between software engineering theory and applied machine learning pipeline creation. It comprises two main artifacts:

1. The Formal Report: A granular, code-level analysis tracking line-by-line engineering choices within Scikit-learn's underlying core system.
2. The Custom Extension: A working, production-grade Python script implementing a custom transformer, classifier, and pipeline ecosystem that inherits from and complies with standard Scikit-learn Mixins.

The custom framework developed within this project explicitly showcases the four pillars of Object-Oriented Programming:

Both SmartClassifier and SmartTransformer inherit concurrently from Scikit-Learn's structural layout.

• SmartClassifier uses BaseEstimator and ClassifierMixin to automatically gain hyperparameter utilities and uniform validation properties.
• SmartTransformer leverages TransformerMixin to acquire automatic fit_transform() behavior through boilerplate code reuse.

Standard behavioral workflows (fit, predict, transform, score) are overridden to execute custom math patterns while maintaining plug-and-play compatibility with native estimators.

Granular mathematical manipulations (such as data matrix centering, state constraints tracking via check_array, and multi-class bound logic parsing) are encapsulated away from the terminal client layer behind generic interface loops.

Rather than using heavy, deep structural inheritance trees, SmartPipeline implements a strict Composition Pattern. It acts as an independent execution orchestrator that holds and controls standalone component objects (SmartTransformer and SmartClassifier), respecting the software axiom: "Favor object composition over class inheritance".

Below is the accurate deployment structure of our project directories:

SCIKIT-LEARN-OOP-ANALYSIS/
│
├── analyzed_files/       # Contains raw/processed data files used during system validation
├── Code/                 # Core Python engine containing the custom_extension.py script
├── diagrams/             # Visual UML, inheritance trees, and pipeline architecture diagrams
├── Report/               # Formal academic technical analysis report document
│
├── .gitignore            # Python environment and metadata runtime filter file
└── README.md             # Project Blueprint & System Deployment Manual (This File)