Update app.py

app.py

@@ -3,9 +3,10 @@ import pickle
 import numpy as np
 
 # Modeli yükle
-with open("model/xgboost_optimized_v2.pkl", "rb") as f:
+with open("xgboost_optimized_v2.pkl", "rb") as f:  # model/ klasörü kaldırıldı
     model_pkg = pickle.load(f)
 
+# Modeller ve fonksiyonlar
 xgb_improved = model_pkg["base_models"]["xgboost_improved"][0]
 smart_align = model_pkg["enhancement_functions"]["smart_feature_alignment"]
 ensemble_fn = model_pkg["enhancement_functions"]["ensemble_predict"]
@@ -13,33 +14,38 @@ unknown_boost_fn = model_pkg["enhancement_functions"]["unknown_category_boost"]
 dynamic_adjust_fn = model_pkg["enhancement_functions"]["dynamic_risk_adjustment"]
 confidence_fn = model_pkg["enhancement_functions"]["calculate_confidence"]
 
+# Ön işleme bileşenleri
 preprocessors = model_pkg["preprocessors"]
 scaler = preprocessors["scaler"]
 encoders = preprocessors["encoders"]
 bert_extractor = preprocessors["bert_extractor"]
 label_encoder = preprocessors["label_encoder"]
-feature_names = preprocessors["scaler"].mean_.shape[0]
+feature_names = scaler.mean_.shape[0]  # Özellik sayısı
 
+# Tahmin fonksiyonu
 def predict(cve_id, cwe_id, description, cvss_score):
-    # Dummy features for now – Hugging Face Spaces does not allow GPU by default
+    # Dummy girişler (GPU olmadığı için BERT çıkarımı yapılmıyor)
     numeric = np.zeros((1, feature_names))
-    bert_feats = np.zeros((1, 768))  # Dummy
+    bert_feats = np.zeros((1, 768))  # Dummy BERT vektörü
     x_input = np.hstack([numeric, bert_feats])
 
+    # Model tahminleri
     xgb_pred = xgb_improved.predict(x_input)[0]
     ensemble_pred = ensemble_fn(cvss_score, xgb_pred)
     final_pred = unknown_boost_fn(ensemble_pred, unknown_features=2, cvss_score=cvss_score)
     optimized = dynamic_adjust_fn(final_pred, cvss_score, cwe_id, len(description))
     confidence = confidence_fn(2, cvss_score)
 
+    # Kategori sınıflandırması
     category = "Düşük" if optimized <= 3 else "Orta" if optimized <= 7 else "Yüksek"
 
     return {
         "Tahmini Risk Skoru": round(optimized, 2),
         "Kategori": category,
-        "Model Güveni": f"%{round(confidence*100, 1)}"
+        "Model Güveni": f"%{round(confidence * 100, 1)}"
     }
 
+# Gradio arayüzü
 interface = gr.Interface(
     fn=predict,
     inputs=[
@@ -53,4 +59,5 @@ interface = gr.Interface(
     description="CVE tanımlarına göre tahmini risk skoru ve kategori"
 )
 
+# Uygulamayı başlat
 interface.launch()