Neural Network for Multi-Class Classification

This repository contains the implementation of a neural network designed for multi-class classification, developed as part of the "Web and Information Retrieval" course at the Università degli Studi del Sannio.

Table of Contents

• Project Overview
• Repository Structure
• Requirements
• Usage
  • 1. Clone the Repository
  • 2. Run the Jupyter Notebook (Optional)
  • 3. Execute the Main Script
• Model Architecture
• Results
• License
• Acknowledgments

Project Overview

The project focuses on building a neural network with a softmax activation function in the final layer, enabling it to handle multi-class classification tasks effectively. The implementation includes data preprocessing, model training, evaluation, and visualization of results.

Repository Structure

Web-Inf/
├── main.py                  # Main script for model training and evaluation
├── tesina_finale_softmax.ipynb  # Detailed notebook explaining the project workflow
├── resources/              # Directory containing datasets and additional resources
├── results/                # Directory to store training results and figures
├── .gitignore              # Files and folders to be ignored by git
├── LICENSE                 # Project license information
└── README.md               # Project documentation

Requirements

To run the code in this repository, ensure you have the following dependencies installed:

• Python 3.x
• TensorFlow
• NumPy
• Pandas
• scikit-learn
• Matplotlib
• Jupyter Notebook (optional, for exploring the notebook)

You can install the required packages using pip:

pip install tensorflow numpy pandas scikit-learn matplotlib jupyter

Usage

1. Clone the Repository

git clone https://github.com/sroman0/Web-Inf.git
cd Web-Inf

2. Run the Jupyter Notebook (Optional)

To explore the project workflow in an interactive environment:

jupyter notebook tesina_finale_softmax.ipynb

3. Execute the Main Script

To train and evaluate the model, run:

python main.py

Ensure that the dataset is correctly referenced within the script and that the resources/ directory contains the necessary data files.

Model Architecture

The implemented neural network consists of:

1. Input Layer: Processes preprocessed feature vectors.
2. Hidden Layers: Multiple dense layers with ReLU activation functions to capture complex patterns.
3. Output Layer: Softmax activation function for multi-class classification.

The network is optimized using the Adam optimizer and uses categorical cross-entropy as the loss function. Model performance is tracked using accuracy metrics.

Results

• Accuracy & Loss Plots: Training history is visualized to assess model convergence.
• Confusion Matrix: Provides insights into class-wise prediction accuracy.
• Evaluation Metrics: Detailed performance metrics, including precision, recall, and F1-score.
• All generated results and plots are saved in the results/ directory.

License

This project is licensed under the GNU General Public License v3.0.
See the LICENSE file for more details.

Acknowledgments

This project was developed as part of the "Web and Information Retrieval" course at Università degli Studi del Sannio. Special thanks to the course instructors and fellow students for their valuable guidance and support.

---

For questions or further information, please contact the repository maintainer or refer to the course materials provided by the university.