fix

up

.env

@@ -1,9 +1,11 @@
 # MySQL Database Configuration
+DB_CONNECTION=mysql
 DB_HOST=localhost
 DB_PORT=3306
-DB_USER=kali
-DB_PASSWORD=asu
+DB_USER=root
+DB_PASSWORD=
 DB_NAME=deteksi_pmk
 
 # Flask Configuration
-FLASK_SECRET=deteksi-pmk-secret-key-2026
+FLASK_SECRET=deteksi-pmk-secret-key-2026
+FLASK_DEBUG=1

.gitignore

@@ -5,6 +5,9 @@ models/*
 !models/scaler.pkl
 !models/label_encoder.pkl
 !models/pca.pkl
+!models/multiclass_knn_model.pkl
+!models/multiclass_scaler.pkl
+!models/multiclass_label_encoder.pkl
 results/*
 uploads/*
 .venv

AGENTS.md

@@ -77,7 +77,7 @@ deteksi_PMK/
 │   ├── __init__.py
 │   ├── mysql_db.py            # SQLAlchemy ORM — models, CRUD, seed data (916 baris)
 │   ├── preprocessing.py       # Validasi sapi + preprocessing pipeline (340 baris)
-│   ├── feature_extraction.py   # Ekstraksi fitur: RGB avg + HSV + GLCM + histogram + Hu (46 fitur)
+│   ├── feature_extraction.py   # Ekstraksi fitur: RGB avg + HSV + GLCM + histogram + Hu (44 fitur)
 │   └── helpers.py             # Load/save model, dataset prep, confidence estimation (237 baris)
 │
 ├── templates/                 # Jinja2 templates
@@ -121,7 +121,7 @@ User upload gambar
   │
   ▼
 [feature_extraction.py] FeatureExtractor.extract_all_features()
-  → 46 fitur: [RGB avg (3), HSV mean+std (6), GLCM (6), histogram 24, Hu (7)]
+  → 44 fitur: [RGB avg (3), HSV mean+std (6), GLCM (4), histogram 24, Hu (7)]
   │
   ▼
 [helpers.py] scaler.transform() → [BINARY MODEL] predict()
@@ -232,7 +232,7 @@ expert_diseases
 -- Aturan forward chaining (5 aturan)
 expert_rules
   id              INT AUTO_INCREMENT PK
-  code            VARCHAR(20) UNIQUE -- FC01–FC04
+  code            VARCHAR(20) UNIQUE -- FC01–FC05
   symptom_codes   TEXT               -- JSON array of symptom codes
   result_disease_code VARCHAR(10) FK → expert_diseases.code
   description     TEXT
@@ -262,12 +262,12 @@ expert_rules_expert_symptoms
 - P03 = PMK_LAKTASI (gejala ambing)
 - P05 = PMK_AKUT_GENERAL (gejala umum berat)
 
-**4 Aturan (FC01–FC04):**
+**5 Aturan (FC01–FC05):**
 
 - FC01 → P01 (oral): [G01, G02, G03, G04, G11, G18, G19, G20, G21]
 - FC02 → P02 (podal): [G01, G02, G05, G06, G15, G22, G23, G24, G25]
 - FC03 → P03 (laktasi): [G01, G02, G07, G08, G09, G26, G27]
-- FC04 → P05 (akut umum): [G01, G02, G03, G04, G05, G06, G07, G09, G11, G12, G14, G18, G20, G22, G23, G24, G26]
+- FC05 → P05 (akut umum): [G01, G02, G03, G04, G05, G06, G07, G09, G11, G12, G14, G18, G20, G22, G23, G24, G26]
 
 > **Auto-check gejala dari hasil deteksi gambar** (di `app.py:pmk_to_symptoms`):
 > Mapping ini digunakan saat redirect ke expert system via mode=image. Hanya gejala spesifik body part yang diikutkan (tanpa gejala umum):
@@ -310,7 +310,7 @@ expert_rules_expert_symptoms
 **Arsitektur Hierarchical (2-level):**
 
 ```
-Input image → 46 fitur (RGB+HSV+GLCM+Histogram+Hu)
+Input image → 44 fitur (RGB+HSV+GLCM+Histogram+Hu)
                    │
             ┌──────┴──────┐
             ▼              ▼
@@ -350,12 +350,12 @@ Input image → 46 fitur (RGB+HSV+GLCM+Histogram+Hu)
 
 | Parameter         | Value                                                                       |
 | ----------------- | --------------------------------------------------------------------------- |
-| **Features**      | 46: avg_red, avg_green, avg_blue, hsv_h_mean, hsv_h_std, hsv_s_mean, hsv_s_std, hsv_v_mean, hsv_v_std, contrast, homogeneity, correlation, energy, dissimilarity, asm, 8-bin histogram × 3 channels (24), hu_moment_0..6 (7) |
+| **Features**      | 44: avg_red, avg_green, avg_blue, hsv_h_mean, hsv_h_std, hsv_s_mean, hsv_s_std, hsv_v_mean, hsv_v_std, contrast, homogeneity, correlation, energy, 8-bin histogram × 3 channels (24), hu_moment_0..6 (7) |
 | **Image size**      | 256×256                                                                     |
 | **Preprocessing** | Otsu threshold, mask, resize                                                |
 | **Scaling**       | StandardScaler                                                              |
 | **Output**        | Binary: `sehat` / `sakit`; Multi-class: `pmk_oral`, `pmk_podal`, `pmk_laktasi`, `pmk_akut_general`, ... |
-| **Confidence**    | Custom weighted neighbor, clipped 50.0–98.5%                                |
+| **Confidence**    | Custom weighted neighbor, clipped 50.0–89.5%                                |
 | **Class detection** | Auto-detect all folders under `dataset/` — `healthy` → `sehat`, `pmk_*` → nama folder |
 
 > **Jika menambah/mengubah fitur**, update:

app.py

@@ -13,7 +13,7 @@ from dotenv import load_dotenv
 load_dotenv()
 
 from utils.helpers import load_model, estimate_prediction_confidence
-from utils.preprocessing import preprocess_image, preprocess_pipeline, validate_cattle_image
+from utils.preprocessing import preprocess_image, preprocess_pipeline, validate_cattle_image, detect_udder
 from utils.feature_extraction import FeatureExtractor
 from expert_system import ForwardChaining, KnowledgeBase  # Import sistem pakar
 
@@ -422,6 +422,20 @@ def _predict_single(filepath, original_filename):
         multi_enc = multiclass_model.predict(multi_scaled)[0]
         pmk_type = multiclass_label_encoder.inverse_transform([multi_enc])[0]
 
+    # Fallback heuristic: jika model menandai 'sehat' tapi gambar mengandung ciri ambing/puting,
+    # anggap sebagai laktasi (pmk_laktasi) dan ubah prediksi agar sistem pakar terpicu.
+    if prediction.lower() == 'sehat':
+        try:
+            udder_detected, udder_score = detect_udder(filepath)
+            if udder_detected:
+                prediction = 'sakit'
+                pmk_type = 'pmk_laktasi'
+                # update confidence minimal berdasarkan score heuristik
+                confidence = max(confidence, min(udder_score * 100.0, 95.0))
+                print(f"[HEURISTIC] Udder heuristic triggered for {original_filename} | score={udder_score:.2f}")
+        except Exception as e:
+            print(f"[HEURISTIC] Udder heuristic error: {e}")
+
     features_dict = {name: float(features[i]) for i, name in enumerate(extractor.feature_names)}
     features_table = list(zip(extractor.feature_names, [round(float(x), 4) for x in features]))
 
@@ -584,6 +598,10 @@ def predict():
     pmk_to_symptoms = {
         'pmk_oral': ['G02', 'G03', 'G04', 'G10', 'G14', 'G18', 'G19', 'G20', 'G21'],
         'pmk_podal': ['G05', 'G06', 'G15', 'G22', 'G23', 'G24'],
+        # Udder (laktasi) — gunakan hanya gejala spesifik ambing/puting
+        'pmk_laktasi': ['G07', 'G08', 'G09', 'G26', 'G27'],
+        # General akut: kombinasi gejala yang menunjukkan PMK akut/menular luas
+        'pmk_akut_general': ['G01', 'G02', 'G03', 'G04', 'G05', 'G06', 'G07', 'G09', 'G11', 'G12', 'G14', 'G18', 'G20', 'G22', 'G23', 'G24', 'G26']
     }
 
     preselected = set()
@@ -592,13 +610,24 @@ def predict():
         if key in pmk_to_symptoms:
             preselected.update(pmk_to_symptoms[key])
 
+    # Jika lebih dari satu body-part berbeda terdeteksi sakit (misal: mulut+kaki, mulut+puting, kaki+puting),
+    # anggap kemungkinan PMK akut/general dan preselect gejala umum akut juga.
+    body_part_keys = {k for k in sick_pmk_types if k in {'pmk_oral', 'pmk_podal', 'pmk_laktasi'}}
+    if len(body_part_keys) >= 2:
+        # tambahkan gejala PMK akut general
+        preselected.update(pmk_to_symptoms.get('pmk_akut_general', []))
+        # juga tambahkan tipe pmk_akut_general ke hasil sehingga UI/riwayat menampilkan tipe ini
+        sick_pmk_types.add('pmk_akut_general')
+
     # Store all images in session for expert system display
     session['last_prediction'] = {
         'db_id': pred_id,
         'filename': upload_images[0]['filename'],
         'original_filename': upload_images[0]['original_filename'],
+        'filepath': upload_images[0]['filepath'],
         'prediction': 'sakit' if sick_pmk_types else 'sehat',
         'confidence': upload_images[0]['confidence'],
+        'features_table': upload_images[0]['features_table'],
         'source': 'image_processing',
     }
     session['upload_images_data'] = upload_images
@@ -809,7 +838,7 @@ def expert_system_page():
         
         # Jika ada hasil prediksi berbasis image processing sebelumnya, tambahkan ke konteks
         last_prediction = session.get('last_prediction', {})
-        upload_images = session.get('upload_images_data', [])
+        upload_images = session.get('upload_images_data', []) if use_image_context else []
         image_info = None
         if use_image_context and last_prediction.get('source') == 'image_processing':
             image_info = {
@@ -837,7 +866,7 @@ def expert_system_page():
 
     # Ambil informasi gambar dari session jika hasil sebelumnya berasal dari image processing
     last_prediction = session.get('last_prediction', {})
-    upload_images = session.get('upload_images_data', [])
+    upload_images = session.get('upload_images_data', []) if use_image_context else []
     image_info = None
     if use_image_context and last_prediction.get('source') == 'image_processing':
         image_info = {