diff --git a/images/accuracy.png b/images/accuracy.png
new file mode 100644
index 0000000..ef0b6b5
Binary files /dev/null and b/images/accuracy.png differ
diff --git a/main.py b/main.py
index 344faea..5cef138 100644
--- a/main.py
+++ b/main.py
@@ -25,29 +25,35 @@ def main():
     X, y = separate_features_and_target(df_clean)
     
     # split into train and test data
-    X_train, X_test, y_train, y_test = train_test_split(
-        X, y, test_size=0.2, stratify=y, random_state=0
-    )
+    accuracy_scores = []
+    # run training many times using different splits to get an average accuracy score 
+    for i in range(1000):
+        X_train, X_test, y_train, y_test = train_test_split(
+            X, y, test_size=0.2, stratify=y, random_state=i
+        )
     
-    # pre training processing
-    X_train_normalized, X_test_normalized = normalize_features(X_train, X_test)
-    
-    # training
-    model = train_logistic_regression(X_train_normalized, y_train)
-    
-    # prediction
-    y_pred = predict_target(model, X_test_normalized)
+        # pre training processing
+        X_train_normalized, X_test_normalized = normalize_features(X_train, X_test)
+        
+        # training
+        model = train_logistic_regression(X_train_normalized, y_train)
+        
+        # prediction
+        y_pred = predict_target(model, X_test_normalized)
 
-    # evaluation
-    le = get_label_encoder(df_clean)
-    # draw_feature_importance(model, X)
-    # draw_confusion_matrix(y_test, y_pred, le)
-    # draw_classification_report(y_test, y_pred, le)
-    evaluate_accuracy(y_test, y_pred)
+        # evaluation
+        le = get_label_encoder(df_clean)
+        # draw_feature_importance(model, X)
+        # draw_confusion_matrix(y_test, y_pred, le)
+        # draw_classification_report(y_test, y_pred, le)
+        accuracy = get_accuracy(y_test, y_pred)
+        accuracy_scores.append(accuracy)
+    print(f"Average Accuracy: {np.mean(accuracy_scores):.4f}")
+    print(f"Samples: {len(accuracy_scores)}")
 
-def evaluate_accuracy(y_test, y_pred):
-    acc = accuracy_score(y_test, y_pred)
-    print(f"Model Accuracy: {acc:.4f}")
+def get_accuracy(y_test, y_pred):
+    accuracy = accuracy_score(y_test, y_pred)
+    return accuracy
 
 def get_label_encoder(df):
     le = LabelEncoder()
diff --git a/readme.md b/readme.md
index e28c2a6..180c48a 100644
--- a/readme.md
+++ b/readme.md
@@ -46,7 +46,7 @@ No missing values or duplicate rows were found in the dataset. Outliers in numer
 
 ![Missing Values](images/missing_values.png)
 ![Duplicate Entries](images/duplicate_entries.png)
-![Duplicate Entries](images/removed_outliers.png)
+![Removed Outliers](images/removed_outliers.png)
 
 ---
 
@@ -60,4 +60,12 @@ To improve model performance and reduce redundancy, I performed feature engineer
 
 ## Modeling
 
-This model was made using **logistic regression**, it works well in this situation because it models the probability of each class based on the input features, making it effective for categorical outcomes. After experimenting with different hyperparameter settings, including various solvers and iteration limits, I found that removing them entirely did not noticeably change the model's performance, indicating that the default configuration worked well enough for this purpose.
\ No newline at end of file
+This model was made using **logistic regression**, it works well in this situation because it models the probability of each class based on the input features, making it effective for categorical outcomes. After experimenting with different hyperparameter settings, including various solvers and iteration limits, I found that removing them entirely did not noticeably change the model's performance, indicating that the default configuration worked well enough for this purpose.
+
+---
+
+## Results
+
+![Accuracy](images/accuracy.png)
+![Classification Report](images/classification_report.png)
+![Confusion Matrix](images/confusion_matrix.png)
\ No newline at end of file