🌐 Feature-Based Spectrum Sensing in NOMA for Cognitive IoT Networks with Optimal ML Classifiers

This project dives into Rayleigh fading channels for Non-Orthogonal Multiple Access (NOMA) users. 📡 Leveraging Monte Carlo simulations, it explores signal detection across different Signal-to-Noise Ratios (SNRs). 📊 Plus, we harness the power of Machine Learning to boost signal detection capabilities.

📚 Libraries and Dependencies

• numpy: For number crunching 🔢.
• pandas: Expert at data handling and manipulation 📈.
• matplotlib: Our go-to for stunning visualizations and plots 📊.
• scipy: A powerhouse for scientific computing, essential for interpolation 🔬.
• scikit-learn: The brain behind our Machine Learning models 🤖.

🔢 Constants and Parameters

• Monte Carlo iterations: num_iter = 10000
• NOMA users: N = 2
• Power allocations: a1, a2 = 0.8, 0.2
• Sampled length: S = 4096
• False-alarm probability: Pf = 0.1
• Environmental SNR range: -25dB to 5dB 🔉.
• Transmitter power: transmitter_power = 1

📡 Data Generation & Signal Detection

Utilizing the Monte Carlo simulation, we:

1. Craft NOMA signals with random cyclic delays 🔄.
2. Merge these to form our transmitted signal 📶.
3. Stir in Rayleigh distributed noise for realism 🌪️.
4. Detect NOMA signals with cyclic correlation 🔍.

🤖 Machine Learning Insights

We train three ML champions:

• Logistic Regression (LR)
• Random Forest (RF)
• Decision Tree (DT)

They're on a mission to spot NOMA signals. We assess their prowess using the ROC curve, eyeing the True Positive Rate (TPR) against a False Positive Rate (FPR) of 0.1.

📊 Visual Mastery

We plot Detection Probability vs. Environmental SNR, showcasing:

• Classic signal detection (sans ML) 🚦
• Logistic Regression (LR) 📈
• Random Forest (RF) 🌳
• Decision Tree (DT) 🌲

🚀 Running the Show

1. Clone our repository using git clone https://github.com/suyashvsingh/NOMA-ML-Spectrum-Detection-CIoT.git 📂.
2. Install the dependencies using pip install -r requirements.txt 📦.
3. Fire up the notebook for an epic plot showdown: traditional vs ML-augmented detection under varied SNRs 📉.

📈 The Results

A stunning graph, "Probability of Detection vs. Environmental SNR," awaits you. It pits traditional detection against our ML trio, LR, RF, and DT, focusing on TPR at a steady FPR of 0.1.

🤝 Join the Mission

Jump into:

• Turbocharge the Monte Carlo simulation ⚙️.
• Bring in cutting-edge ML algorithms 🧠.
• Elevate our visualization game 🎨.

---

Note: For the whole story and deep insights, explore the notebook 📘.