🧠 Slime Mould Algorithm for Breast Cancer Disease Subtype Classification Using Multiomics Data

This project presents an intelligent web-based tool for classifying breast cancer subtypes based on multiomics data using advanced optimization algorithms. The core objective is to enhance classification performance through Slime Mould Algorithm (SMA)-based feature selection, and compare its performance against Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) using Random Forest and SVM classifiers.

---

📁 Project Structure

BCD Care/
│
├── app.py                           # Main Flask application
│
├── models/                          # Model scripts and saved .pkl artifacts
│   ├── randomforest_withsma.py
│   ├── randomforest_withga.py
│   ├── randomforest_withpso.py
│   ├── svm_withsma.py
│   ├── svm_withga.py
│   ├── svm_withpso.py
│   ├── *.pkl                        # Trained models and selected feature files
│   └── report.log                   # Logging information
│
├── project/uploads/                # Dataset storage
│   └── brca_data_w_subtypes (1).csv
│
├── static/
│   ├── *.png                        # Logos and image assets
│   ├── css/
│   │   ├── detection.css
│   │   ├── eda.css
│   │   ├── navbar.css
│   │   ├── resources.css
│   │   └── style.css
│   └── js/
│       ├── detection.js
│       └── eda.js
│
├── templates/
│   ├── index.html
│   ├── detection.html
│   ├── eda.html
│   └── resources.html
│
└── uploads/
    └── brca_data_w_subtypes (1).csv  # Duplicate for runtime

---

🔬 Project Overview

• Objective: Classify patients' breast cancer subtypes (Alive/Dead) using multiomics data.
• Core Technique: Slime Mould Algorithm (SMA) for biologically-inspired feature selection.
• Comparative Methods: Genetic Algorithm (GA), Particle Swarm Optimization (PSO).
• Classifiers: Random Forest, SVM (Support Vector Machine).
• Balancing Method: SMOTE (Synthetic Minority Oversampling Technique).

---

💻 Features

• 🔍 Upload and analyze multiomics datasets in .csv format.
• 📊 Perform EDA with visual feature correlations.
• ⚙️ Select from six different model pipelines:
  • Random Forest with SMA, GA, PSO
  • SVM with SMA, GA, PSO
• 📈 Visual performance metrics and comparison:
  • ROC Curve
  • Confusion Matrix
  • Convergence Curves
  • Radar Metrics
  • Class Distribution (SMOTE)

---

⚙️ Setup Instructions

1. Clone and Install

git clone https://github.com/adwityac/bcd-care.git
cd bcd-care
pip install -r requirements.txt

Sample requirements.txt

Flask
Flask-CORS
numpy
pandas
scikit-learn
matplotlib
seaborn
joblib
imblearn
lightgbm
deap
pyswarms

---

🚀 How to Run

python app.py

Visit: http://127.0.0.1:5000

---

🧪 How It Works

1. Data Upload:

Upload your dataset (must contain vital.status column).

2. Model Selection:

Choose a model from dropdown:

• SMA, GA, PSO (RF or SVM variant)

3. Result:

• Prediction (Alive/Dead)
• Metric outputs (Accuracy, F1, AUC, etc.)
• Performance charts
• SMOTE visualization

---

📊 Evaluation Metrics

• Accuracy
• Precision / Recall / F1-Score
• ROC AUC
• Log Loss
• Sensitivity & Specificity
• Radar Chart comparing all

---

📈 Visual Outputs

• Feature Correlation Heatmap (EDA)
• Confusion Matrix
• ROC Curve
• Log Loss and Accuracy trends
• Radar Performance Chart
• SMA / PSO Convergence Plot

---

🎓 Authors

This project was developed as part of academic research at Banasthali Vidyapith:

• Ashmita Mukherjee
• Adwitya Chakraborty
• Kriti
• Kajal

Supervisor: Dr. Rahul Kumar Vijay

---

📜 License

This project is licensed under the MIT License.