Data Science x Logistic Regression: A Magical Classifier

Introduction: A Muggle’s Mission at Hogwarts

In this project, inspired by the world of Harry Potter, we took on the role of data scientists tasked with creating a machine learning model to replace the malfunctioning Sorting Hat. Using logistic regression, we must classify students into the four Hogwarts houses based on their academic performance. Objectives: Learning the Fundamentals of Logistic Regression

This assignment serves as an introduction to classification problems, focusing on:

• Data Exploration – Understanding, analyzing, and cleaning datasets.
• Data Visualization – Using histograms, scatter plots, and pair plots to identify patterns.
• Logistic Regression – Implementing a multi-class classification model using the one-vs-all strategy.
• Model Training & Prediction – Writing custom code to train a model using gradient descent and make predictions.

Core Tasks: Steps to Build the Sorting Algorithm

1. Data Analysis

Before building a model, understanding the dataset is crucial. We were required to:

• Examine the dataset’s structure.
• Compute basic statistical properties (count, mean, std, min, max) without using pre-built functions like Pandas’ describe().

2. Data Visualization

Visualizing data helps in feature selection and detecting anomalies. We must

• Histograms to assess score distributions across
• Scatter plots to compare features and detect
• Pair plots to analyze relationships between multiple features.

3. Logistic Regression

The heart of the project is implementing logistic regression for classification. We had to:

Train a one-vs-all classifier using gradient descent to optimize weights.
Create two programs:
    logreg_train to train the model and store weights.
    logreg_predict to classify new students and generate a prediction file.

Bonus: Enhancing the Model

• Implementing stochastic gradient descent or other optimization techniques.
• Expanding statistical analysis with more descriptive features.