QUICKSTART GUIDE — How to Run the Neuro-Fuzzy Project

This guide shows a person how to run all commands without confusion. Just copy-paste the commands into your terminal.

---

overview of the training results for DUVAL, ROGERS AND DRM:

📁 Project Structure

After running predictions, all outputs are saved to:

fuzzy_logic/
├── outputs/                    # All prediction & comparison results
│   ├── anfis/                  # ANFIS model outputs
│   ├── random_forest/          # Random Forest model outputs
│   └── comparison/             # Combined model comparisons & metrics
├── randomforest/               # Random Forest module (do not edit)
├── results/                    # Trained models & encoders (do not edit)
└── [data files]

---

🚀 Quick Commands

1️⃣ Run ANFIS Model Only

All three methods (Duval, Rogers, DRM):

cd c:\Users\USER\my_desktop\neuro_fuzzy\fuzzy_logic
python predict_fault.py --file test_model.csv --method all

Single method (e.g., Duval):

python predict_fault.py --file test_model.csv --method duval

Single sample (gas readings):

python predict_fault.py --single "{\"H2\":100, \"CH4\":50, \"C2H6\":30, \"C2H4\":20, \"C2H2\":5, \"CO\":15}"

Output location: outputs/anfis/

---

2️⃣ Run Random Forest Model Only

All three methods:

python randomforest/rf_predict.py --file test_model.csv --method all

Single method (e.g., Rogers):

python randomforest/rf_predict.py --file test_model.csv --method rogers

Single sample:

python randomforest/rf_predict.py --single "{\"H2\":100, \"CH4\":50, \"C2H6\":30, \"C2H4\":20, \"C2H2\":5, \"CO\":15}"

Output location: outputs/random_forest/

---

3️⃣ Compare Both Models (ANFIS + Random Forest)

Compare all methods on a CSV file:

python compare_models.py --file test_model.csv --method all --output compare_results.csv

Compare with ground-truth labels (if you have them):

python compare_models.py --file labeled_test.csv --method all --label-column FAULT --output results.csv

Compare single samples:

python compare_models.py --single "{\"H2\":100, \"CH4\":50, \"C2H6\":30, \"C2H4\":20, \"C2H2\":5, \"CO\":15}" --method duval

Output location: outputs/comparison/

• results.csv — merged predictions from both models
• results_metrics.txt — accuracy, precision, recall, F1-scores
• cm_*.png — confusion matrix plots
• agreement_report.txt — how often ANFIS & RF agree

---

📊 What Each Command Does

Command	Input	Output	When to Use
predict_fault.py --file X.csv	CSV with gas readings	ANFIS predictions	Quick ANFIS-only predictions
randomforest/rf_predict.py --file X.csv	CSV with gas readings	RF predictions	Quick RF-only predictions
compare_models.py --file X.csv	CSV with gas readings	Both models' predictions + metrics	Compare ANFIS vs RF
compare_models.py --file X.csv --label-column FAULT	CSV with gas readings + ground truth	Full evaluation report + confusion matrices	Full evaluation & analysis

---

🎯 Example Workflows

Workflow 1: Test ANFIS on your data

cd c:\Users\USER\my_desktop\neuro_fuzzy\fuzzy_logic
python predict_fault.py --file your_data.csv --method duval
# Check outputs/anfis/ for results

Workflow 2: Test Random Forest on your data

cd c:\Users\USER\my_desktop\neuro_fuzzy\fuzzy_logic
python randomforest/rf_predict.py --file your_data.csv --method duval
# Check outputs/random_forest/ for results

Workflow 3: Compare both models and get full metrics (with labeled data)

cd c:\Users\USER\my_desktop\neuro_fuzzy\fuzzy_logic
python compare_models.py --file your_labeled_data.csv --label-column TrueLabel --method all
# Check outputs/comparison/ for:
#   - results.csv (merged predictions)
#   - results_metrics.txt (accuracy, precision, recall, F1)
#   - cm_*.png (confusion matrix plots)

---

📝 Input File Format

Your CSV file should contain these columns (in any order):

• H2 — Hydrogen concentration (ppm)
• CH4 — Methane concentration (ppm)
• C2H6 — Ethane concentration (ppm)
• C2H4 — Ethylene concentration (ppm)
• C2H2 — Acetylene concentration (ppm)
• CO — Carbon Monoxide concentration (ppm)

Optional: If comparing with ground truth, add a label column:

• FAULT or TrueLabel or ground_truth (name doesn't matter, just specify it with --label-column)

Example labeled file:

H2,CH4,C2H6,C2H4,C2H2,CO,FAULT
100,50,30,20,5,15,Normal
500,200,100,80,10,50,T2
1000,800,300,200,50,100,T3

---

🔧 Advanced Usage

Single Method Only

# Run only Duval method (ANFIS)
python predict_fault.py --file test.csv --method duval

# Run only Rogers method (Random Forest)
python randomforest/rf_predict.py --file test.csv --method rogers

Custom JSON Sample Input

# For Duval method (4 features: H2, CH4, C2H6, C2H4)
python predict_fault.py --single "{\"H2\":100, \"CH4\":50, \"C2H6\":30, \"C2H4\":20}"

# For Rogers method (6 features: all)
python compare_models.py --single "{\"H2\":100, \"CH4\":50, \"C2H6\":30, \"C2H4\":20, \"C2H2\":5, \"CO\":15}" --method rogers

---

📂 Output Directory Structure

After running predictions, you'll see:

outputs/
├── anfis/
│   ├── duval_predictions.csv
│   ├── rogers_predictions.csv
│   ├── drm_predictions.csv
│   └── anfis_summary.txt
├── random_forest/
│   ├── duval_rf_predictions.csv
│   ├── rogers_rf_predictions.csv
│   ├── drm_rf_predictions.csv
│   └── rf_summary.txt
└── comparison/
    ├── results.csv
    ├── results_metrics.txt
    ├── cm_duval_anfis.png
    ├── cm_duval_rf.png
    ├── cm_rogers_anfis.png
    ├── cm_rogers_rf.png
    ├── cm_drm_anfis.png
    ├── cm_drm_rf.png
    └── agreement_report.txt

---

✅ Troubleshooting

"File not found" error

• Make sure your CSV file is in the fuzzy_logic/ directory, or provide the full path:

  python predict_fault.py --file "C:\path\to\your\file.csv"

"Column not found" error

• Your CSV must have the required gas concentration columns: H2, CH4, C2H6, C2H4, C2H2, CO
• Check that column names match exactly (case-sensitive)

No metrics printed

• Make sure you provided --label-column with the correct column name
• Example: --label-column FAULT (if your CSV has a column named "FAULT")

Many warnings about "StandardScaler feature names"

• These are non-fatal warnings — predictions are still correct
• Warnings go away in updated version (auto-fixed by refactoring)

---

📞 Summary

Goal	Command
Quick test of ANFIS	python predict_fault.py --file test_model.csv --method all
Quick test of RF	python randomforest/rf_predict.py --file test_model.csv --method all
Compare both models	python compare_models.py --file test_model.csv --method all
Full evaluation (with labels)	python compare_models.py --file labeled.csv --label-column FAULT --method all
Test single sample	python compare_models.py --single "{...}" --method duval

All outputs go to outputs/ directory. No messy files scattered around!

---

For questions or issues, check the command syntax above or see individual module READMEs.