dslr

📝 Overview

• DSLR (Data Science Logistic Regression) is a one-vs-all logistic regression classifier designed to predict Hogwarts Houses based on student attributes.
• The project follows gradient descent optimization to train a logistic regression model and make predictions with high accuracy.

The goal is to classify students into Gryffindor, Slytherin, Hufflepuff, or Ravenclaw based on features extracted from the dataset.
The model is evaluated against a ground truth dataset, and its accuracy is measured.

---

🔥 Features

✔️ Train a logistic regression model using batch, stochastic, or mini-batch gradient descent
✔️ Make predictions for new data based on trained weights
✔️ Evaluate accuracy using Scikit-Learn’s accuracy score
✔️ Graphic visualizations like histograms and scatter plots
✔️ Weights saved and loaded from JSON for reuse in predictions

---

📁 Setup

⚙️ Install requirements

This project requires Python and the following libraries:

• numpy
• pandas
• matplotlib

🚀 Usage

1️⃣ Clone this repository:

git clone https://github.com/your-username/dslr.git
cd dslr

2️⃣ Train the Model:

• Train the model on dataset_train.csv and save the weights to weights.json:

python3 logreg_train.py data/dataset_train.csv

• Optional: Use Stochastic gradient descent (-s) or Mini-batch (-m)

python3 logreg_train.py datasets/dataset_train.csv -s

2️⃣ Predict Hogwarts Houses Use the trained model to classify students from dataset_test.csv:

python3 logreg_predict.py datasets/dataset_test.csv weights.json

3️⃣ Evaluate the Model

python3 evaluate.py

📊 Visualize the Data

Explore the dataset using different visualizations:

📌 Histogram: Shows the distribution of numerical features.

python3 histogram.py

📌 Scatter Plot: Displays feature relationships across Hogwarts Houses.

python3 scatter_plot.py

📌 Pair Plot: Helps identify patterns and separability between classes.

python3 pair_plot.py

✨ My Teammate => Belen