5G Network Slice Classification Using CNN

This project demonstrates the use of Convolutional Neural Networks (CNNs) to classify 5G network slices based on simulated data. It represents a practical application of deep learning in the domain of next-generation wireless networks.

Slice Categories

The classification focuses on three core 5G functionalities:

• eMBB – Enhanced Mobile Broadband
• mMTC – Massive Machine-Type Communications
• URLLC – Ultra-Reliable Low-Latency Communications

Features

🔬 Synthetic Data Simulation

• Simulates 5G network metrics such as latency, throughput, signal strength, jitter, etc.
• Each sample is labeled with a slice category based on these metrics.

CNN Model

• A custom-built CNN model is trained to classify slices based on the synthetic data.
• Optimized for performance using TensorFlow/Keras.

Performance Visualization

• Training Accuracy & Loss Curves
• Confusion Matrix for classification breakdown
• ROC Curves with class-wise AUC scores
• Pair Plot with KDE for feature distribution analysis
• Simulated Heatmap showing resource utilization trends across slices

🛠️ Technologies Used

• Programming Language: Python
• Libraries:
  • TensorFlow / Keras
  • Matplotlib
  • Seaborn
  • Scikit-learn

📁 File Structure

• 5G_Slice_Classification.ipynb – Google Colab notebook containing:
  • Synthetic data generation
  • CNN model definition and training
  • Visual evaluation of the results

How to Run

1. Download or Clone this repository.
2. Open the Notebook (5G_Slice_Classification.ipynb) in Google Colab.
3. Run all cells sequentially to:
   • Generate synthetic dataset
   • Train the CNN
   • Visualize model outputs

📈 Visual Outputs

• Training Accuracy and Loss: Understand how the model learns over time.
• Confusion Matrix: Evaluate classification performance across eMBB, mMTC, URLLC.
• ROC Curves: Measure true-positive vs. false-positive rate for each class.
• Pair Plot with KDE: Visualize feature relationships and distributions.
• Simulated Heatmap: Display 5G slice resource usage trends.

👥 Contributors

This project was collaboratively developed by:

• Harshit Soni
• Hridyansh Sharma
• Shreyas Makwana
• Daksh Mehta

We worked together to conceptualize, develop, and present this project, combining our unique skills to build an end-to-end deep learning pipeline for 5G slice classification.

---

📌 Note: This project is built on simulated data and intended for academic demonstration purposes only.