🥗 SmartFoodSelector

---

🍴 Make Smarter Food Choices with AI

SmartFoodSelector revolutionizes food analysis and recommendation using cutting-edge AI! 🚀 This modular system combines machine learning, Bayesian inference, and symbolic logic to:

• 🧹 Preprocess and normalize nutritional data.
• 🎯 Cluster products into meaningful groups via k-Means.
• 🤖 Predict categories for new items with Random Forests and Logistic Regression.
• 🔮 Model probabilistic relationships with Bayesian Networks.
• 🧩 Execute logical queries via Prolog knowledge bases.

---

🌟 Key Features

🧹 1. Data Preprocessing

• Cleaning: Handle missing values and outliers.
• Feature Selection: Focus on core nutritional metrics (energy_100g, fat_100g, etc.).
• Normalization: Scale features using MinMaxScaler for balanced analysis.

🎯 2. Unsupervised Clustering (k-Means)

• Elbow Method: Automatically determine optimal clusters with kneed.
• Visualization: Explore cluster distributions via pie charts and PCA-reduced plots.
• Output: Generate clustered_dataset.csv for downstream tasks.

🤖 3. Supervised Learning

• Models: Train Decision Trees, Random Forests, and Logistic Regression.
• Balancing: Address class imbalance with SMOTE.
• Evaluation: Compare metrics (Accuracy, F1-score, Precision/Recall) and interpret results via SHAP values.
• Learning Curves: Diagnose overfitting/underfitting.

🔮 4. Bayesian Networks

• Continuous & Discrete Models: Learn probabilistic dependencies with pgmpy.
• Inference: Predict preferences and handle missing data.
• Visualization: Plot dependency graphs for interpretability.

🧩 5. Prolog Knowledge Base

• Automated Generation: Convert clustered data into Prolog facts/rules.
• Query Interface: Use pyswip to execute logical rules like:
  • product_info(E, F, C, Su, P, Sa, Cluster) for cluster lookup.
  • Custom constraints (e.g., high_protein_low_sugar).

---

🛠️ Architecture

src/
├── dataset_preprocessing.py    # Data cleaning/normalization
├── unsupervised_clustering.py  # k-Means + Elbow Method
├── supervised_trainer.py       # Model training/evaluation
├── bayes_net.py                # Bayesian Networks
├── prolog_interface.py         # Prolog query handler
└── generate_prolog_knowledge_base.py  # Prolog KB generator
main.py                         # Pipeline coordinator

---

📦 Get Started

1. Install Dependencies:

   pip install pandas scikit-learn pgmpy pyswip optuna shap kneed

2. Preprocess Data:

   python main.py --preprocess

3. Run Clustering:

   python main.py --cluster

4. Train Models:

   python main.py --train

5. Launch Prolog KB:

   python main.py --prolog

---

📊 Results & Observations

• Clustering: Optimal k=3 clusters identified via elbow method.
• Supervised Models: Random Forest achieved highest accuracy (92%).
• Bayesian Networks: Enabled probabilistic inference under missing data.
• Prolog Integration: Efficiently answered complex nutritional constraints.

---

🤝 Contribute

• Report issues or suggest enhancements via GitHub Issues.
• Submit Pull Requests for bug fixes/new features.

---

📜 License

MIT License. See LICENSE for details.

---

🌟 Empower your food decisions with AI and logic! 🌟

For full technical details, refer to the project documentation.

---

Key Updates:

• Added architectural overview and setup instructions.
• Expanded technical details while retaining engaging tone.
• Linked to full documentation for deeper exploration.
• Simplified CLI commands for ease of use.