5G ML Prototype

Streamlit-based 5G machine learning prototype with two workflows:

1. Performance Prediction (pretrained inference)
2. ML Lab (train/compare models, feature selection, correlation analysis, optional LSTM benchmark)

Additional capabilities:

• Precise optimization strategy generation with target-based actions
• SHAP explainability plots (optional)
• Research-grade diagnostics (cross-validation, timing, residual/confusion analysis, ROC-AUC indicators)

Project Layout

.
├── app.py
├── requirements.txt
├── analysis/
│   ├── correlation_graphs.py
│   ├── explainability.py
│   ├── feature_selection.py
│   ├── optimization_strategies.py
│   └── research_plots.py
├── models/
│   ├── model_registry.py
│   ├── lstm_model.py
│   ├── linear_regression_model.py
│   ├── logistic_regression_model.py
│   ├── random_forest_regressor_model.py
│   ├── random_forest_classifier_model.py
│   ├── gradient_boosting_regressor_model.py
│   ├── gradient_boosting_classifier_model.py
│   ├── svr_model.py
│   ├── svc_model.py
│   ├── knn_regressor_model.py
│   ├── knn_classifier_model.py
│   ├── mlp_regressor_model.py
│   └── mlp_classifier_model.py
├── data/
│   ├── 5g_network_data.csv
│   └── sample_input.csv
├── artifacts/
│   └── pretrained/
│       ├── latency_model.pkl
│       ├── throughput_model.pkl
│       └── qos_model.pkl
└── docs/
		├── QUICKSTART.md
		└── SYSTEM_EXPLANATION.md

Setup

Install dependencies:

pip install -r requirements.txt

Run app:

streamlit run app.py

Are the PKL files required?

Short answer: depends on tab.

• Required for Performance Prediction tab:
  • artifacts/pretrained/latency_model.pkl
  • artifacts/pretrained/throughput_model.pkl
  • artifacts/pretrained/qos_model.pkl
• Not required for ML Lab tab, because ML Lab trains models from data/5g_network_data.csv or uploaded CSV.

If you only use ML Lab, you can keep the pretrained files archived elsewhere. If you use Prediction, keep them in artifacts/pretrained/.

Documentation

• Quick start: docs/QUICKSTART.md
• Full architecture, diagrams, and flowcharts: docs/SYSTEM_EXPLANATION.md

Latest Research Features

The Streamlit ML Lab now includes additional research outputs for thesis writing:

1. Dataset quality and numeric statistics tables.
2. Strongest feature-correlation table.
3. Target-specific correlation ranking.
4. Telecom KPI relationship plots.
5. Optional default-vs-tuned model comparison using GridSearchCV.
6. Exportable experiment package for chapter tables and screenshots.

Recommended thesis workflow:

1. Open the ML Lab tab.
2. Select the target, such as latency, throughput, or qos_class.
3. Enable Research Mode.
4. Run cross-validation and tuned comparison.
5. Download the leaderboard, correlations, diagnostics, and experiment summary.
6. Use these exports in the Methodology, Experimental Results, and Discussion chapters.