🧠 C-Accelerated Perceptron for Binary Classification ⚙️

This project implements a Perceptron algorithm for binary classification, with its main design functions written in C for performance optimization and wrapped in a Python class for easy use, data manipulation, and visualization. 📊 The project includes functionality to use the Iris dataset (converted to a binary problem) or generate synthetic data for training and testing.

🎯 Goal: Create an efficient Perceptron classifier with a C backend and a friendly Python interface.

✨ Example View:

🌟 Overview

Perceptron is one of the simplest supervised machine learning algorithms for binary classification. This project demonstrates:

• Implementation of the Perceptron algorithm.
• Usage of C for computationally intensive operations (neuron calculation, weight adjustment, accuracy evaluation) via ctypes in Python.
• A Python Perceptron class that encapsulates the training, prediction, evaluation, and plotting logic.
• Loading and preprocessing data for the Iris dataset and generating synthetic data. 🌸
• Splitting data into training and testing sets.
• Feature scaling (Z-score normalization).
• Training with a defined learning rate and accuracy threshold.
• Calculating error, accuracy, and recall. 📈
• K-fold cross-validation. * Visualization of:
• Training and testing data 📍
• Decision boundary of the trained model 🗺️
• Model accuracy over epochs 🎯
• Evolution of weights during training 🏋️
• Model error over epochs 📉

✨ Main Features

• C Core ⚙️: neuron, fit, evaluate_accuracy, predict and recall functions, innovations in C for efficiency.
• Python Wrapper 🐍: Easy-to-use Perceptron class in Python.
• Data Sources 💾:
• Utilize the Iris dataset (filtered for two classes and two features).
• Synthetic data generation for classification problems (linearly separable or noisy).
• Preprocessing 🧹:
• Convert multiclass problems to binary.
• Scale features using mean and standard deviation.
• Training 🏋️‍♀️:
• Iterate until a baseline accuracy is reached on the test set or a maximum number of epochs is reached.
• Store historical weights, errors, and accuracy.
• Evaluation 📊:
• Calculate accuracy on the training and test sets.
• Perform k-fold cross-validation.
• Visualization 🖼️: Use matplotlib to plot:
• Data distribution.
• Decision boundary.
• Learning curves (accuracy, error, weights).
• Analysis 🖼️: * Flexibility 🛠️: Allows configuration of learning rate, reference accuracy and data generation parameters.

📂 Project Structure

├── 📄 perceptron.c # C implementation of the main Perceptron functions

├── 📄 perceptron.h # Header file for the C code

├── 🔗 perceptron.so # Compiled shared library (generated after compilation)

├── 🐍 data_treatment.py # Functions for data import, generation and plotting

├── 🐍 training.py # Python Perceptron class and ctypes interface for C

└── 📖 main.py # Main script to run training and evaluation

🛠️ Prerequisites

• Python 3.12.3 🐍
• C compiler (like CGC) ⚙️
• Python libraries:
• numpy
• matplotlib
• scikit-learn (used in data_treatment.py for load_iris and make_classification)

🚀 Setup and Installation

1. Clone the repository:

git clone https://github.com/JvFg92/Perceptron_Data_Classify
cd Perceptron_Data_Classify

2. Compile the C code to create a shared library (perceptron.so): On Linux or macOS:

gcc -shared -o perceptron.so -fPIC perceptron.c

On Windows (may require configuration depending on your compiler, e.g. with MinGW):

gcc -shared -o perceptron.so perceptron.c -Wl,--add-stdcall-alias

ℹ️ Make sure the resulting perceptron.so (or perceptron.dll on Windows) file is in the same directory as the Python scripts.

3. Install Python dependencies:

pip install numpy matplotlib scikit-learn

⚠️ For Linux you may need:

sudo apt install python3-numpy
sudo apt install python3-matplotlib
sudo apt install python3-sklearn

⚠️ For Windows you may need:

py -m pip install numpy matplotlib scikit-learn

✅ Ready to go!

▶️ Usage

The main script to run the model is main.py.

python main.py

⚠️ For Linux you may need:

python3 main.py

⚠️ For Windows you may need:

py main.py