remove random state, matrix labels on specific rows

models/X_train_sample.npy

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dacc23f02577d428bb68c9f15def2373ee2dfac9ed3e55f4f328add123f54384
-size 11027608
+oid sha256:6c297b12b2ad73e2c32805b123ab456d0c57135c33469314cbb570588a653f1f
+size 12393352

models/scaler.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27ae467c5aa5c69269cb2c2524518b1985eda83b9bed02ebf0fd1ad16e61d8a2
+oid sha256:c7c5bd15219def3155b39a51b6627cf54d82b87acc63de2be03e7e3073793435
 size 1487

models/skin_cancer_model.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7867c457b12b6d092ccadcb34d0c6c2921620b2416f6f1e53b925a0ad24d700c
-size 118655649
+oid sha256:e815f34d3e4a2538499199c8eb4529d3b8719263b736727ea1c992941c287dc8
+size 135524049

src/data.py

@@ -25,8 +25,7 @@ def balance_dataset(df):
         df_resampled = resample(
             class_subset,
             replace=True,
-            n_samples=max_count,
-            random_state=42
+            n_samples=max_count
         )
         balanced_dfs.append(df_resampled)
 
@@ -34,7 +33,7 @@ def balance_dataset(df):
     df_balanced = pd.concat(balanced_dfs)
 
     # Shuffle the dataset so classes aren't grouped together
-    df_balanced = df_balanced.sample(frac=1, random_state=42).reset_index(drop=True)
+    df_balanced = df_balanced.sample(frac=1).reset_index(drop=True)
 
     print(f"Original size: {len(df)} -> Balanced size: {len(df_balanced)}")
     return df_balanced
@@ -62,7 +61,7 @@ def load_metadata(limit=None, balance=True):
     if limit:
         print(f"Subsampling to {limit}...")
         actual_limit = min(limit, len(df))
-        df, _ = train_test_split(df, train_size=actual_limit, stratify=df['label_idx'], random_state=42)
+        df, _ = train_test_split(df, train_size=actual_limit, stratify=df['label_idx'])
 
     # Apply balancing (Upsampling)
     # if balance:

src/explainability.py

@@ -67,10 +67,22 @@ def generate_lime_explanations(model, X_train, X_test, y_test, feature_names, cl
         )
 
         # 4. Save Plot
-        # We title it with the True Label for context
-        true_label = class_names[y_test[i]]
-        fig = exp.as_pyplot_figure()
-        plt.title(f"LIME ({model_name}): Test Instance {i} | True Label: {true_label}")
+        # FIX: We must explicitly tell pyplot which label to plot.
+        # exp.local_exp keys are the class indices that were explained.
+        # Since we used top_labels=1, there is only one key.
+        available_labels = list(exp.local_exp.keys())
+        if not available_labels:
+            continue
+
+        explained_label_idx = available_labels[0]
+
+        # Get class name for title
+        pred_label_name = class_names[explained_label_idx]
+        true_label_name = class_names[y_test[i]]
+
+        # Pass the specific label we computed to avoid KeyError
+        fig = exp.as_pyplot_figure(label=explained_label_idx)
+        plt.title(f"LIME ({model_name}): Test Inst {i}\nTrue: {true_label_name} | Pred: {pred_label_name}")
         plt.tight_layout()
 
         save_path = os.path.join(output_dir, f'{model_name}_inst_{i}_lime.png')

src/model.py

@@ -18,8 +18,8 @@ def train_and_evaluate_split(X_train, y_train, X_test, y_test, classes):
 
     # 1. Define Models
     techniques = {
-        "RF": RandomForestClassifier(n_estimators=100, class_weight='balanced', random_state=42),
-        "SVM": SVC(probability=True, class_weight='balanced', random_state=42)
+        "RF": RandomForestClassifier(n_estimators=100, class_weight='balanced'),
+        "SVM": SVC(probability=True, class_weight='balanced')
     }
 
     best_score = 0

src/plots.py

@@ -10,23 +10,31 @@ from . import config
 def plot_confusion_matrix(y_true, y_pred, classes, model_name):
     """Generates and saves both Raw and Normalized confusion matrix heatmaps."""
 
-    # 1. Calculate Raw Matrix
-    cm = confusion_matrix(y_true, y_pred)
+    # 1. Determine Alphabetical Order
+    # y_true/y_pred are indices (0, 1, 2...) mapping to the original 'classes' list.
+    # We want to display them in alphabetical order of the class names.
+    sorted_indices = np.argsort(classes)
+    sorted_classes = np.array(classes)[sorted_indices]
+
+    # 2. Calculate Raw Matrix
+    # passing 'labels=sorted_indices' forces the matrix rows/cols to follow the alphabetical order
+    cm = confusion_matrix(y_true, y_pred, labels=sorted_indices)
 
     # Plot Raw Counts
     plt.figure(figsize=(10, 8))
     sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
-                xticklabels=classes, yticklabels=classes)
+                xticklabels=sorted_classes, yticklabels=sorted_classes)
     plt.title(f"{model_name} Confusion Matrix (Counts)")
     plt.ylabel('True Label')
     plt.xlabel('Predicted Label')
+    plt.xticks(rotation=45)
     plt.tight_layout()
 
     filename = f"{model_name.lower()}_confusion_matrix.png"
     plt.savefig(os.path.join(config.MODEL_DIR, filename))
     plt.close()
 
-    # 2. Calculate Normalized Matrix
+    # 3. Calculate Normalized Matrix
     # Divide each row element by the sum of that row (True Label count)
     cm_norm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
     # Replace NaN with 0 (safe guard for empty classes)
@@ -36,10 +44,11 @@ def plot_confusion_matrix(y_true, y_pred, classes, model_name):
     plt.figure(figsize=(10, 8))
     # Use fmt='.2f' to show 2 decimal places (e.g., 0.95)
     sns.heatmap(cm_norm, annot=True, fmt='.2f', cmap='Greens',
-                xticklabels=classes, yticklabels=classes)
+                xticklabels=sorted_classes, yticklabels=sorted_classes)
     plt.title(f"{model_name} Confusion Matrix (Normalized)")
     plt.ylabel('True Label')
     plt.xlabel('Predicted Label')
+    plt.xticks(rotation=45)
     plt.tight_layout()
 
     filename_norm = f"{model_name.lower()}_confusion_matrix_normalized.png"

train_main.py

@@ -68,7 +68,7 @@ def main():
 
     # Split DataFrame FIRST to avoid data leakage
     df_train, df_test = train_test_split(
-        df, test_size=0.2, stratify=df['label_idx'], random_state=42
+        df, test_size=0.1, stratify=df['label_idx']
     )
 
     print(f"Training Samples (Files): {len(df_train)}")