Update app.py

app.py

@@ -36,77 +36,76 @@ def create_plots(df, feature_columns, target_column):
             # Add target to each feature set
             features = features + [target_column]
             
-            # Create pair plot (scatter)
+            # Create scatter plot
+            plt.figure(figsize=(12, 10))
             if is_numeric_target:
-                pair_plot = sns.pairplot(df[features], kind='scatter', 
-                                         plot_kws={'alpha': 0.6})
+                scatter_plot = sns.pairplot(df[features], kind='scatter', 
+                                            plot_kws={'alpha': 0.6}, corner=True)
                 norm = plt.Normalize(df[target_column].min(), df[target_column].max())
-                for i, ax in enumerate(pair_plot.axes.flatten()):
-                    if i < len(features) * (len(features) - 1):  # Check if it's not a diagonal plot
-                        x_var = features[i // len(features)]
-                        y_var = features[i % len(features)]
-                        if x_var != y_var:
-                            scatter = ax.scatter(df[x_var], df[y_var], c=df[target_column], cmap='viridis', norm=norm)
-                            fig = ax.figure
-                            fig.colorbar(scatter, ax=ax, label=target_column)
+                for ax in scatter_plot.axes.flatten():
+                    if ax.get_xlabel() != ax.get_ylabel():
+                        scatter = ax.collections[0]
+                        scatter.set_cmap('viridis')
+                        scatter.set_norm(norm)
+                        scatter.set_array(df[target_column])
+                        plt.colorbar(scatter, ax=ax, label=target_column)
             else:
-                pair_plot = sns.pairplot(df[features], hue=target_column, kind='scatter')
+                scatter_plot = sns.pairplot(df[features], hue=target_column, kind='scatter', corner=True)
             
-            pair_plot.fig.suptitle(f'Features Distributions - Group {group} (Scatter)', y=1.02)
+            scatter_plot.fig.suptitle(f'Scatter Plots - Group {group}', y=1.02, fontsize=16)
             
             # Adjust label size and spacing
-            for ax in pair_plot.axes.flatten():
-                ax.tick_params(labelsize=8)
-                ax.set_xlabel(ax.get_xlabel(), fontsize=10)
-                ax.set_ylabel(ax.get_ylabel(), fontsize=10)
+            for ax in scatter_plot.axes.flatten():
+                ax.tick_params(labelsize=10)
+                ax.set_xlabel(ax.get_xlabel(), fontsize=12)
+                ax.set_ylabel(ax.get_ylabel(), fontsize=12)
             
-            pair_plot.fig.tight_layout()
-            plt.subplots_adjust(top=0.95, bottom=0.1, left=0.1, right=0.9, hspace=0.5, wspace=0.5)
+            plt.tight_layout()
             
             buf = io.BytesIO()
-            pair_plot.savefig(buf, format='png', dpi=300)
+            plt.savefig(buf, format='png', dpi=300)
             buf.seek(0)
             plots.append(buf)
-            plt.close(pair_plot.fig)
+            plt.close()
 
-            # Create pair plot (histogram)
+            # Create histogram plot
+            plt.figure(figsize=(12, 10))
             if is_numeric_target:
-                pair_plot_hist = sns.pairplot(df[features], kind='hist', 
-                                              plot_kws={'alpha': 0.6})
-                for i, ax in enumerate(pair_plot_hist.axes.flatten()):
-                    x_var = features[i // len(features)]
-                    y_var = features[i % len(features)]
-                    if x_var == y_var:
+                hist_plot = sns.pairplot(df[features], kind='hist', 
+                                         plot_kws={'alpha': 0.6}, corner=True)
+                for ax in hist_plot.axes.flatten():
+                    if ax.get_xlabel() == ax.get_ylabel():
                         ax.clear()
-                        sns.histplot(df[x_var], ax=ax, kde=True)
+                        sns.histplot(df[ax.get_xlabel()], ax=ax, kde=True)
                     else:
-                        scatter = ax.scatter(df[x_var], df[y_var], c=df[target_column], cmap='viridis', norm=norm)
-                        fig = ax.figure
-                        fig.colorbar(scatter, ax=ax, label=target_column)
+                        scatter = ax.collections[0]
+                        scatter.set_cmap('viridis')
+                        scatter.set_norm(norm)
+                        scatter.set_array(df[target_column])
+                        plt.colorbar(scatter, ax=ax, label=target_column)
             else:
-                pair_plot_hist = sns.pairplot(df[features], kind='hist', hue=target_column)
+                hist_plot = sns.pairplot(df[features], kind='hist', hue=target_column, corner=True)
             
-            pair_plot_hist.fig.suptitle(f'Features Distributions - Group {group} (Histogram)', y=1.02)
+            hist_plot.fig.suptitle(f'Histogram Plots - Group {group}', y=1.02, fontsize=16)
             
             # Adjust label size and spacing
-            for ax in pair_plot_hist.axes.flatten():
-                ax.tick_params(labelsize=8)
-                ax.set_xlabel(ax.get_xlabel(), fontsize=10)
-                ax.set_ylabel(ax.get_ylabel(), fontsize=10)
+            for ax in hist_plot.axes.flatten():
+                ax.tick_params(labelsize=10)
+                ax.set_xlabel(ax.get_xlabel(), fontsize=12)
+                ax.set_ylabel(ax.get_ylabel(), fontsize=12)
             
-            pair_plot_hist.fig.tight_layout()
-            plt.subplots_adjust(top=0.95, bottom=0.1, left=0.1, right=0.9, hspace=0.5, wspace=0.5)
+            plt.tight_layout()
             
             buf = io.BytesIO()
-            pair_plot_hist.savefig(buf, format='png', dpi=300)
+            plt.savefig(buf, format='png', dpi=300)
             buf.seek(0)
             plots.append(buf)
-            plt.close(pair_plot_hist.fig)
+            plt.close()
 
-            # Create regplot grid
+            # Create regression plot
             n_features = len(features) - 1  # Exclude target column
-            fig, axes = plt.subplots(n_features, n_features, figsize=(20, 20))
-            fig.suptitle(f'Regression Plots - Group {group}', y=1.02)
+            fig, axes = plt.subplots(n_features, n_features, figsize=(16, 14))
+            fig.suptitle(f'Regression Plots - Group {group}', y=1.02, fontsize=16)
             
             for i, feature1 in enumerate(features[:-1]):
                 for j, feature2 in enumerate(features[:-1]):
@@ -121,23 +120,22 @@ def create_plots(df, feature_columns, target_column):
                             plt.colorbar(scatter, ax=ax, label=target_column)
                         else:
                             sns.regplot(x=feature1, y=feature2, data=df, ax=ax, 
-                                        line_kws={"color": "black"})
+                                        scatter_kws={'alpha': 0.6}, line_kws={'color': 'red'})
                     else:
                         sns.histplot(df[feature1], ax=ax, kde=True)
                     
-                    ax.set_xlabel(feature1 if i == n_features - 1 else '')
-                    ax.set_ylabel(feature2 if j == 0 else '')
+                    ax.set_xlabel(feature1, fontsize=10)
+                    ax.set_ylabel(feature2, fontsize=10)
                     ax.tick_params(labelsize=8)
-                    ax.set_title(f'{feature1} vs {feature2}', fontsize=10)
+                    ax.set_title(f'{feature1} vs {feature2}', fontsize=12)
             
-            fig.tight_layout()
-            plt.subplots_adjust(top=0.95, bottom=0.1, left=0.1, right=0.9, hspace=0.5, wspace=0.5)
+            plt.tight_layout()
             
             buf = io.BytesIO()
             plt.savefig(buf, format='png', dpi=300)
             buf.seek(0)
             plots.append(buf)
-            plt.close(fig)
+            plt.close()
         
         # Calculate Pearson correlation values
         correlation_matrix = df[feature_columns + [target_column]].corr()
@@ -213,16 +211,16 @@ with gr.Blocks() as iface:
     analyze_btn = gr.Button("Analyze")
 
     with gr.Row():
-        plot1 = gr.Image(label="Pair Plot (Scatter) - Group 1")
-        plot4 = gr.Image(label="Pair Plot (Scatter) - Group 2")
+        plot1 = gr.Image(label="Scatter Plots - Group 1")
+        plot4 = gr.Image(label="Scatter Plots - Group 2")
 
     with gr.Row():
-        plot2 = gr.Image(label="Pair Plot (Histogram) - Group 1")
-        plot5 = gr.Image(label="Pair Plot (Histogram) - Group 2")
+        plot2 = gr.Image(label="Histogram Plots - Group 1")
+        plot5 = gr.Image(label="Histogram Plots - Group 2")
         
     with gr.Row():
-        plot3 = gr.Image(label="Regression Plot - Group 1")
-        plot6 = gr.Image(label="Regression Plot - Group 2")
+        plot3 = gr.Image(label="Regression Plots - Group 1")
+        plot6 = gr.Image(label="Regression Plots - Group 2")
         
     with gr.Row():
         heatmap = gr.Image(label="Pearson Correlation Heatmap")