fix(viz_reports): plot_scatter_true_vs_pred NaN 容错 + subplots_adjust 替换 tight_layout

src/postprocessing/visualization_reports.py

@@ -81,7 +81,11 @@ class WaterQualityVisualization:
             保存的文件路径
         """
         fig, ax = plt.subplots(figsize=(10, 8))
-        
+
+        # 强制处理 NaN
+        y_true = np.nan_to_num(y_true, nan=0.0)
+        y_pred = np.nan_to_num(y_pred, nan=0.0)
+
         # 计算所有数据的R²和RMSE
         from sklearn.metrics import r2_score, mean_squared_error, mean_absolute_error
         
@@ -133,7 +137,7 @@ class WaterQualityVisualization:
         ax.set_xlabel(f'真实值 ({target_name})', fontsize=14, fontweight='bold')
         ax.set_ylabel(f'预测值 ({target_name})', fontsize=14, fontweight='bold')
         ax.set_title(f'{target_name} - 真实值 vs 预测值', fontsize=16, fontweight='bold')
-        ax.legend(loc='upper left', fontsize=11, bbox_to_anchor=(1.02, 1), borderaxespad=0)
+        ax.legend(loc='upper left', fontsize=11, bbox_to_anchor=(1.05, 1), borderaxespad=0)
         ax.grid(True, alpha=0.3)
         
         # 添加指标文本框
@@ -141,7 +145,7 @@ class WaterQualityVisualization:
                verticalalignment='top', bbox=dict(boxstyle='round', 
                facecolor='wheat', alpha=0.8), fontsize=10)
         
-        plt.tight_layout()
+        plt.subplots_adjust(right=0.85)
         
         # 保存图片
         if output_path is None: