Update app.py

app.py

@@ -155,9 +155,12 @@ def analyze_data(csv_file, additional_notes=""):
     return format_analysis_report(analysis_result, visuals)
 
 def compare_models():
+    import seaborn as sns
+    from sklearn.model_selection import cross_val_predict
+
     if df_global is None:
-        return "Please upload and preprocess a dataset first."
-    
+        return pd.DataFrame({"Error": ["Please upload and preprocess a dataset first."]}), None
+
     target = df_global.columns[-1]
     X = df_global.drop(target, axis=1)
     y = df_global[target]
@@ -168,21 +171,43 @@ def compare_models():
     models = {
         "RandomForest": RandomForestClassifier(),
         "LogisticRegression": LogisticRegression(max_iter=1000),
-        "SVC": SVC()
+        "GradientBoosting": GradientBoostingClassifier()
     }
 
     results = []
     for name, model in models.items():
+        # Cross-validation scores
         scores = cross_val_score(model, X, y, cv=5)
-        results.append({
+
+        # Cross-validated predictions for metrics
+        y_pred = cross_val_predict(model, X, y, cv=5)
+
+        metrics = {
             "Model": name,
             "CV Mean Accuracy": np.mean(scores),
-            "CV Std Dev": np.std(scores)
-        })
-        wandb.log({f"{name}_cv_mean": np.mean(scores), f"{name}_cv_std": np.std(scores)})
+            "CV Std Dev": np.std(scores),
+            "F1 Score": f1_score(y, y_pred, average="weighted", zero_division=0),
+            "Precision": precision_score(y, y_pred, average="weighted", zero_division=0),
+            "Recall": recall_score(y, y_pred, average="weighted", zero_division=0),
+        }
+        wandb.log({f"{name}_{k.replace(' ', '_').lower()}": v for k, v in metrics.items() if isinstance(v, (float, int))})
+        results.append(metrics)
 
     results_df = pd.DataFrame(results)
-    return results_df
+
+    # Plotting
+    plt.figure(figsize=(8, 5))
+    sns.barplot(data=results_df, x="Model", y="CV Mean Accuracy", palette="Blues_d")
+    plt.title("Model Comparison (CV Mean Accuracy)")
+    plt.ylim(0, 1)
+    plt.tight_layout()
+
+    plot_path = "./model_comparison.png"
+    plt.savefig(plot_path)
+    plt.close()
+
+    return results_df, plot_path
+
     
 # 1. prepare_data should come first
 def prepare_data(df, target_column=None):
@@ -352,8 +377,15 @@ with gr.Blocks() as demo:
         shap_img = gr.Image(label="SHAP Summary Plot")
         lime_img = gr.Image(label="LIME Explanation")
 
+    with gr.Row():
+        compare_btn = gr.Button("Compare Models (A/B Testing)")
+        compare_output = gr.DataFrame(label="Model Comparison (CV + Metrics)")
+        compare_img = gr.Image(label="Model Accuracy Plot")
+
     agent_btn.click(fn=analyze_data, inputs=[file_input], outputs=[insights_output, visual_output])
     train_btn.click(fn=train_model, inputs=[file_input], outputs=[metrics_output, trials_output])
     explain_btn.click(fn=explainability, inputs=[], outputs=[shap_img, lime_img])
+    compare_btn.click(fn=compare_models, inputs=[], outputs=[compare_output, compare_img])
+
 
 demo.launch(debug=True)