dragon-ml-toolbox 2.3.0__py3-none-any.whl → 3.0.0__py3-none-any.whl

ml_tools/MICE_imputation.py

@@ -6,6 +6,7 @@ import numpy as np
 from .utilities import load_dataframe, list_csv_paths, sanitize_filename, _script_info, merge_dataframes, save_dataframe, threshold_binary_values, make_fullpath
 from plotnine import ggplot, labs, theme, element_blank # type: ignore
 from typing import Optional, Union
+from .logger import _LOGGER
 
 
 __all__ = [
@@ -40,7 +41,9 @@ def apply_mice(df: pd.DataFrame, df_name: str, binary_columns: Optional[list[str
     if binary_columns is not None:
         invalid_binary_columns = set(binary_columns) - set(df.columns)
         if invalid_binary_columns:
-            print(f"⚠️ These 'binary columns' are not in the dataset: {invalid_binary_columns}")
+            _LOGGER.warning(f"⚠️ These 'binary columns' are not in the dataset:")
+            for invalid_binary_col in invalid_binary_columns:
+                print(f"  - {invalid_binary_col}")
         valid_binary_columns = [col for col in binary_columns if col not in invalid_binary_columns]
         for imputed_df in imputed_datasets:
             for binary_column_name in valid_binary_columns:
@@ -125,7 +128,7 @@ def get_convergence_diagnostic(kernel: mf.ImputationKernel, imputed_dataset_name
             plt.savefig(save_path, bbox_inches='tight', format="svg")
             plt.close()
             
-        print(f"{dataset_file_dir} completed.")
+        _LOGGER.info(f"{dataset_file_dir} completed.")
 
 
 # Imputed distributions
@@ -210,7 +213,7 @@ def get_imputed_distributions(kernel: mf.ImputationKernel, df_name: str, root_di
             fig = kernel.plot_imputed_distributions(variables=[feature])
             _process_figure(fig, feature)
 
-    print(f"{local_dir_name} completed.")
+    _LOGGER.info(f"{local_dir_name} completed.")
 
 
 def run_mice_pipeline(df_path_or_dir: Union[str,Path], target_columns: list[str], 
@@ -240,7 +243,8 @@ def run_mice_pipeline(df_path_or_dir: Union[str,Path], target_columns: list[str]
         all_file_paths = list(list_csv_paths(input_path).values())
     
     for df_path in all_file_paths:
-        df, df_name = load_dataframe(df_path=df_path)
+        df: pd.DataFrame
+        df, df_name = load_dataframe(df_path=df_path, kind="pandas") # type: ignore
         
         df, df_targets = _skip_targets(df, target_columns)
         

ml_tools/ML_callbacks.py

@@ -0,0 +1,341 @@
+import numpy as np
+import torch
+from tqdm.auto import tqdm
+from .utilities import make_fullpath, LogKeys
+from .logger import _LOGGER
+from typing import Optional
+
+
+__all__ = [
+    "Callback", 
+    "History", 
+    "TqdmProgressBar",
+    "EarlyStopping",  
+    "ModelCheckpoint",
+    "LRScheduler"
+]
+
+
+class Callback:
+    """
+    Abstract base class used to build new callbacks.
+    
+    The methods of this class are automatically called by the Trainer at different
+    points during training. Subclasses can override these methods to implement
+    custom logic.
+    """
+    def __init__(self):
+        self.trainer = None
+
+    def set_trainer(self, trainer):
+        """This is called by the Trainer to associate itself with the callback."""
+        self.trainer = trainer
+
+    def on_train_begin(self, logs=None):
+        """Called at the beginning of training."""
+        pass
+
+    def on_train_end(self, logs=None):
+        """Called at the end of training."""
+        pass
+
+    def on_epoch_begin(self, epoch, logs=None):
+        """Called at the beginning of an epoch."""
+        pass
+
+    def on_epoch_end(self, epoch, logs=None):
+        """Called at the end of an epoch."""
+        pass
+
+    def on_batch_begin(self, batch, logs=None):
+        """Called at the beginning of a training batch."""
+        pass
+
+    def on_batch_end(self, batch, logs=None):
+        """Called at the end of a training batch."""
+        pass
+
+
+class History(Callback):
+    """
+    Callback that records events into a `history` dictionary.
+    
+    This callback is automatically applied to every MyTrainer model.
+    The `history` attribute is a dictionary mapping metric names (e.g., 'val_loss')
+    to a list of metric values.
+    """
+    def on_train_begin(self, logs=None):
+        # Clear history at the beginning of training
+        self.trainer.history = {} # type: ignore
+
+    def on_epoch_end(self, epoch, logs=None):
+        logs = logs or {}
+        for k, v in logs.items():
+            # Append new log values to the history dictionary
+            self.trainer.history.setdefault(k, []).append(v) # type: ignore
+
+
+class TqdmProgressBar(Callback):
+    """Callback that provides a tqdm progress bar for training."""
+    def __init__(self):
+        self.epoch_bar = None
+        self.batch_bar = None
+
+    def on_train_begin(self, logs=None):
+        self.epochs = self.trainer.epochs # type: ignore
+        self.epoch_bar = tqdm(total=self.epochs, desc="Training Progress")
+
+    def on_epoch_begin(self, epoch, logs=None):
+        total_batches = len(self.trainer.train_loader) # type: ignore
+        self.batch_bar = tqdm(total=total_batches, desc=f"Epoch {epoch}/{self.epochs}", leave=False)
+
+    def on_batch_end(self, batch, logs=None):
+        self.batch_bar.update(1) # type: ignore
+        if logs:
+            self.batch_bar.set_postfix(loss=f"{logs.get(LogKeys.BATCH_LOSS, 0):.4f}") # type: ignore
+
+    def on_epoch_end(self, epoch, logs=None):
+        self.batch_bar.close() # type: ignore
+        self.epoch_bar.update(1) # type: ignore
+        if logs:
+            train_loss_str = f"{logs.get(LogKeys.TRAIN_LOSS, 0):.4f}"
+            val_loss_str = f"{logs.get(LogKeys.VAL_LOSS, 0):.4f}"
+            self.epoch_bar.set_postfix_str(f"Train Loss: {train_loss_str}, Val Loss: {val_loss_str}") # type: ignore
+
+    def on_train_end(self, logs=None):
+        self.epoch_bar.close() # type: ignore
+
+
+class EarlyStopping(Callback):
+    """
+    Stop training when a monitored metric has stopped improving.
+    
+    Args:
+        monitor (str): Quantity to be monitored. Defaults to 'val_loss'.
+        min_delta (float): Minimum change in the monitored quantity to qualify as an improvement.
+        patience (int): Number of epochs with no improvement after which training will be stopped.
+        mode (str): One of {'auto', 'min', 'max'}. In 'min' mode, training will stop when the quantity
+                    monitored has stopped decreasing; in 'max' mode it will stop when the quantity
+                    monitored has stopped increasing; in 'auto' mode, the direction is automatically
+                    inferred from the name of the monitored quantity.
+        verbose (int): Verbosity mode.
+    """
+    def __init__(self, monitor: str=LogKeys.VAL_LOSS, min_delta=0.0, patience=3, mode='auto', verbose=1):
+        super().__init__()
+        self.monitor = monitor
+        self.patience = patience
+        self.min_delta = min_delta
+        self.wait = 0
+        self.stopped_epoch = 0
+        self.verbose = verbose
+        
+        if mode not in ['auto', 'min', 'max']:
+            raise ValueError(f"EarlyStopping mode {mode} is unknown, choose one of ('auto', 'min', 'max')")
+        self.mode = mode
+
+        # Determine the comparison operator based on the mode
+        if self.mode == 'min':
+            self.monitor_op = np.less
+        elif self.mode == 'max':
+            self.monitor_op = np.greater
+        else: # auto mode
+            if 'acc' in self.monitor.lower():
+                self.monitor_op = np.greater
+            else: # Default to min mode for loss or other metrics
+                self.monitor_op = np.less
+        
+        self.best = np.Inf if self.monitor_op == np.less else -np.Inf
+
+    def on_train_begin(self, logs=None):
+        # Reset state at the beginning of training
+        self.wait = 0
+        self.stopped_epoch = 0
+        self.best = np.Inf if self.monitor_op == np.less else -np.Inf
+                    
+    def on_epoch_end(self, epoch, logs=None):
+        current = logs.get(self.monitor) # type: ignore
+        if current is None:
+            return
+
+        # Determine the comparison threshold based on the mode
+        if self.monitor_op == np.less:
+            # For 'min' mode, we need to be smaller than 'best' by at least 'min_delta'
+            # Correct check: current < self.best - self.min_delta
+            is_improvement = self.monitor_op(current, self.best - self.min_delta)
+        else:
+            # For 'max' mode, we need to be greater than 'best' by at least 'min_delta'
+            # Correct check: current > self.best + self.min_delta
+            is_improvement = self.monitor_op(current, self.best + self.min_delta)
+
+        if is_improvement:
+            if self.verbose > 1:
+                _LOGGER.info(f"EarlyStopping: {self.monitor} improved from {self.best:.4f} to {current:.4f}")
+            self.best = current
+            self.wait = 0
+        else:
+            self.wait += 1
+            if self.wait >= self.patience:
+                self.stopped_epoch = epoch
+                self.trainer.stop_training = True # type: ignore
+                if self.verbose > 0:
+                    print("")
+                    _LOGGER.info(f"Epoch {epoch+1}: early stopping after {self.wait} epochs with no improvement.")
+
+
+class ModelCheckpoint(Callback):
+    """
+    Saves the model to a directory with automated filename generation and rotation. The filename includes the epoch and score.
+
+    - If `save_best_only` is True, it saves the single best model, deleting the
+      previous best. 
+    - If `save_best_only` is False, it keeps the 3 most recent checkpoints,
+      deleting the oldest ones automatically.
+
+    Args:
+        save_dir (str): Directory where checkpoint files will be saved.
+        monitor (str): Metric to monitor for `save_best_only=True`.
+        save_best_only (bool): If true, save only the best model.
+        mode (str): One of {'auto', 'min', 'max'}.
+        verbose (int): Verbosity mode.
+    """
+    def __init__(self, save_dir: str, monitor: str = LogKeys.VAL_LOSS,
+                 save_best_only: bool = False, mode: str = 'auto', verbose: int = 1):
+        super().__init__()
+        self.save_dir = make_fullpath(save_dir, make=True)
+        if not self.save_dir.is_dir():
+            _LOGGER.error(f"{save_dir} is not a valid directory.")
+            raise IOError()
+        
+        self.monitor = monitor
+        self.save_best_only = save_best_only
+        self.verbose = verbose
+
+        # State variables to be managed during training
+        self.saved_checkpoints = []
+        self.last_best_filepath = None
+
+        if mode not in ['auto', 'min', 'max']:
+            raise ValueError(f"ModelCheckpoint mode {mode} is unknown.")
+        self.mode = mode
+
+        if self.mode == 'min':
+            self.monitor_op = np.less
+        elif self.mode == 'max':
+            self.monitor_op = np.greater
+        else:
+            self.monitor_op = np.less if 'loss' in self.monitor else np.greater
+        
+        self.best = np.Inf if self.monitor_op == np.less else -np.Inf
+
+    def on_train_begin(self, logs=None):
+        """Reset state when training starts."""
+        self.best = np.Inf if self.monitor_op == np.less else -np.Inf
+        self.saved_checkpoints = []
+        self.last_best_filepath = None
+
+    def on_epoch_end(self, epoch, logs=None):
+        logs = logs or {}
+        self.save_dir.mkdir(parents=True, exist_ok=True)
+
+        if self.save_best_only:
+            self._save_best_model(epoch, logs)
+        else:
+            self._save_rolling_checkpoints(epoch, logs)
+
+    def _save_best_model(self, epoch, logs):
+        """Saves a single best model and deletes the previous one."""
+        current = logs.get(self.monitor)
+        if current is None:
+            return
+
+        if self.monitor_op(current, self.best):
+            old_best_str = f"{self.best:.4f}" if self.best not in [np.Inf, -np.Inf] else "inf"
+            
+            # Create a descriptive filename
+            filename = f"epoch_{epoch}-{self.monitor}_{current:.4f}.pth"
+            new_filepath = self.save_dir / filename
+            
+            if self.verbose > 0:
+                print("")
+                _LOGGER.info(f"Epoch {epoch}: {self.monitor} improved from {old_best_str} to {current:.4f}, saving model to {new_filepath}")
+            
+            # Save the new best model
+            torch.save(self.trainer.model.state_dict(), new_filepath) # type: ignore
+
+            # Delete the old best model file
+            if self.last_best_filepath and self.last_best_filepath.exists():
+                self.last_best_filepath.unlink()
+            
+            # Update state
+            self.best = current
+            self.last_best_filepath = new_filepath
+
+    def _save_rolling_checkpoints(self, epoch, logs):
+        """Saves the latest model and keeps only the last 5."""
+        filename = f"epoch_{epoch}.pth"
+        filepath = self.save_dir / filename
+        
+        if self.verbose > 0:
+            print("")
+            _LOGGER.info(f'Epoch {epoch}: saving model to {filepath}')
+        torch.save(self.trainer.model.state_dict(), filepath) # type: ignore
+
+        self.saved_checkpoints.append(filepath)
+
+        # If we have more than n checkpoints, remove the oldest one
+        if len(self.saved_checkpoints) > 3:
+            file_to_delete = self.saved_checkpoints.pop(0)
+            if file_to_delete.exists():
+                if self.verbose > 0:
+                    _LOGGER.info(f"  -> Deleting old checkpoint: {file_to_delete.name}")
+                file_to_delete.unlink()
+
+
+class LRScheduler(Callback):
+    """
+    Callback to manage a PyTorch learning rate scheduler.
+
+    This callback automatically calls the scheduler's `step()` method at the
+    end of each epoch. It also logs a message when the learning rate changes.
+
+    Args:
+        scheduler: An initialized PyTorch learning rate scheduler.
+        monitor (str, optional): The metric to monitor for schedulers that
+                                 require it, like `ReduceLROnPlateau`.
+                                 Should match a key in the logs (e.g., 'val_loss').
+    """
+    def __init__(self, scheduler, monitor: Optional[str] = None):
+        super().__init__()
+        self.scheduler = scheduler
+        self.monitor = monitor
+        self.previous_lr = None
+
+    def on_train_begin(self, logs=None):
+        """Store the initial learning rate."""
+        self.previous_lr = self.trainer.optimizer.param_groups[0]['lr'] # type: ignore
+
+    def on_epoch_end(self, epoch, logs=None):
+        """Step the scheduler and log any change in learning rate."""
+        # For schedulers that need a metric (e.g., val_loss)
+        if isinstance(self.scheduler, torch.optim.lr_scheduler.ReduceLROnPlateau):
+            if self.monitor is None:
+                raise ValueError("LRScheduler needs a `monitor` metric for ReduceLROnPlateau.")
+            
+            metric_val = logs.get(self.monitor) # type: ignore
+            if metric_val is not None:
+                self.scheduler.step(metric_val)
+            else:
+                print("")
+                _LOGGER.warning(f"LRScheduler could not find metric '{self.monitor}' in logs.")
+        
+        # For all other schedulers
+        else:
+            self.scheduler.step()
+
+        # Log the change if the LR was updated
+        current_lr = self.trainer.optimizer.param_groups[0]['lr'] # type: ignore
+        if current_lr != self.previous_lr:
+            print("")
+            _LOGGER.info(f"Epoch {epoch}: Learning rate changed to {current_lr:.6f}")
+            self.previous_lr = current_lr
+            

ml_tools/ML_evaluation.py

@@ -0,0 +1,255 @@
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.metrics import (
+    classification_report, 
+    ConfusionMatrixDisplay, 
+    roc_curve, 
+    roc_auc_score, 
+    mean_squared_error,
+    mean_absolute_error,
+    r2_score, 
+    median_absolute_error
+)
+import torch
+import shap
+from pathlib import Path
+from .utilities import make_fullpath
+from .logger import _LOGGER
+from typing import Union, Optional
+
+
+__all__ = [
+    "plot_losses", 
+    "classification_metrics", 
+    "regression_metrics",
+    "shap_summary_plot"
+]
+
+
+def plot_losses(history: dict, save_dir: Optional[Union[str, Path]] = None):
+    """
+    Plots training & validation loss curves from a history object.
+
+    Args:
+        history (dict): A dictionary containing 'train_loss' and 'val_loss'.
+        save_dir (str | Path | None): Directory to save the plot image.
+    """
+    train_loss = history.get('train_loss', [])
+    val_loss = history.get('val_loss', [])
+    
+    if not train_loss and not val_loss:
+        print("Warning: Loss history is empty or incomplete. Cannot plot.")
+        return
+
+    fig, ax = plt.subplots(figsize=(10, 5), dpi=100)
+    
+    # Plot training loss only if data for it exists
+    if train_loss:
+        epochs = range(1, len(train_loss) + 1)
+        ax.plot(epochs, train_loss, 'o-', label='Training Loss')
+    
+    # Plot validation loss only if data for it exists
+    if val_loss:
+        epochs = range(1, len(val_loss) + 1)
+        ax.plot(epochs, val_loss, 'o-', label='Validation Loss')
+    
+    ax.set_title('Training and Validation Loss')
+    ax.set_xlabel('Epochs')
+    ax.set_ylabel('Loss')
+    ax.legend()
+    ax.grid(True)
+    plt.tight_layout()
+    
+    if save_dir:
+        save_dir_path = make_fullpath(save_dir, make=True)
+        save_path = save_dir_path / "loss_plot.svg"
+        plt.savefig(save_path)
+        _LOGGER.info(f"Loss plot saved as '{save_path.name}'")
+    else:
+        plt.show()
+    plt.close(fig)
+
+
+def classification_metrics(y_true: np.ndarray, y_pred: np.ndarray, y_prob: Optional[np.ndarray] = None, 
+                           cmap: str = "Blues", save_dir: Optional[Union[str, Path]] = None):
+    """
+    Displays and optionally saves classification metrics and plots.
+
+    Args:
+        y_true (np.ndarray): Ground truth labels.
+        y_pred (np.ndarray): Predicted labels.
+        y_prob (np.ndarray, optional): Predicted probabilities for ROC curve.
+        cmap (str): Colormap for the confusion matrix.
+        save_dir (str | Path | None): Directory to save plots. If None, plots are shown not saved.
+    """
+    print("--- Classification Report ---")
+    report: str = classification_report(y_true, y_pred) # type: ignore
+    print(report)
+    
+    if save_dir:
+        save_dir_path = make_fullpath(save_dir, make=True)
+        # Save text report
+        report_path = save_dir_path / "classification_report.txt"
+        report_path.write_text(report, encoding="utf-8")
+        _LOGGER.info(f"Classification report saved as '{report_path.name}'")
+
+        # Save Confusion Matrix
+        fig_cm, ax_cm = plt.subplots(figsize=(6, 6), dpi=100)
+        ConfusionMatrixDisplay.from_predictions(y_true, y_pred, cmap=cmap, ax=ax_cm)
+        ax_cm.set_title("Confusion Matrix")
+        cm_path = save_dir_path / "confusion_matrix.svg"
+        plt.savefig(cm_path)
+        _LOGGER.info(f"Confusion matrix saved as '{cm_path.name}'")
+        plt.close(fig_cm)
+
+        # Save ROC Curve
+        if y_prob is not None and y_prob.ndim > 1 and y_prob.shape[1] >= 2:
+            fpr, tpr, _ = roc_curve(y_true, y_prob[:, 1])
+            auc = roc_auc_score(y_true, y_prob[:, 1])
+            fig_roc, ax_roc = plt.subplots(figsize=(6, 6), dpi=100)
+            ax_roc.plot(fpr, tpr, label=f'AUC = {auc:.2f}')
+            ax_roc.plot([0, 1], [0, 1], 'k--')
+            ax_roc.set_title('Receiver Operating Characteristic (ROC) Curve')
+            ax_roc.set_xlabel('False Positive Rate')
+            ax_roc.set_ylabel('True Positive Rate')
+            ax_roc.legend(loc='lower right')
+            ax_roc.grid(True)
+            roc_path = save_dir_path / "roc_curve.svg"
+            plt.savefig(roc_path)
+            _LOGGER.info(f"ROC curve saved as '{roc_path.name}'")
+            plt.close(fig_roc)
+    else:
+        # Show plots if not saving
+        ConfusionMatrixDisplay.from_predictions(y_true, y_pred, cmap=cmap)
+        plt.show()
+        if y_prob is not None and y_prob.ndim > 1 and y_prob.shape[1] >= 2:
+            fpr, tpr, _ = roc_curve(y_true, y_prob[:, 1])
+            auc = roc_auc_score(y_true, y_prob[:, 1])
+            plt.figure()
+            plt.plot(fpr, tpr, label=f'AUC = {auc:.2f}')
+            plt.plot([0, 1], [0, 1], 'k--')
+            plt.title('ROC Curve')
+            plt.show()
+
+
+def regression_metrics(y_true: np.ndarray, y_pred: np.ndarray, save_dir: Optional[Union[str, Path]] = None):
+    """
+    Displays regression metrics and optionally saves plots and report.
+
+    Args:
+        y_true (np.ndarray): Ground truth values.
+        y_pred (np.ndarray): Predicted values.
+        save_dir (str | None): Directory to save plots and report.
+    """
+    rmse = np.sqrt(mean_squared_error(y_true, y_pred))
+    mae = mean_absolute_error(y_true, y_pred)
+    r2 = r2_score(y_true, y_pred)
+    medae = median_absolute_error(y_true, y_pred)
+    
+    report_lines = [
+        "--- Regression Report ---",
+        f"  Root Mean Squared Error (RMSE): {rmse:.4f}",
+        f"  Mean Absolute Error (MAE):      {mae:.4f}",
+        f"  Median Absolute Error (MedAE):  {medae:.4f}",
+        f"  Coefficient of Determination (R²): {r2:.4f}"
+    ]
+    report_string = "\n".join(report_lines)
+    print(report_string)
+
+    if save_dir:
+        save_dir_path = make_fullpath(save_dir, make=True)
+        # Save text report
+        report_path = save_dir_path / "regression_report.txt"
+        report_path.write_text(report_string)
+        _LOGGER.info(f"Regression report saved as '{report_path.name}'")
+
+        # Save residual plot
+        residuals = y_true - y_pred
+        fig_res, ax_res = plt.subplots(figsize=(8, 6), dpi=100)
+        ax_res.scatter(y_pred, residuals, alpha=0.6)
+        ax_res.axhline(0, color='red', linestyle='--')
+        ax_res.set_xlabel("Predicted Values")
+        ax_res.set_ylabel("Residuals")
+        ax_res.set_title("Residual Plot")
+        ax_res.grid(True)
+        plt.tight_layout()
+        res_path = save_dir_path / "residual_plot.svg"
+        plt.savefig(res_path)
+        _LOGGER.info(f"Residual plot saved as '{res_path.name}'")
+        plt.close(fig_res)
+
+        # Save true vs predicted plot
+        fig_tvp, ax_tvp = plt.subplots(figsize=(8, 6), dpi=100)
+        ax_tvp.scatter(y_true, y_pred, alpha=0.6)
+        ax_tvp.plot([y_true.min(), y_true.max()], [y_true.min(), y_true.max()], 'k--', lw=2)
+        ax_tvp.set_xlabel('True Values')
+        ax_tvp.set_ylabel('Predictions')
+        ax_tvp.set_title('True vs. Predicted Values')
+        ax_tvp.grid(True)
+        plt.tight_layout()
+        tvp_path = save_dir_path / "true_vs_predicted_plot.svg"
+        plt.savefig(tvp_path)
+        _LOGGER.info(f"True vs. Predicted plot saved as '{tvp_path.name}'")
+        plt.close(fig_tvp)
+
+
+def shap_summary_plot(model, background_data: torch.Tensor, instances_to_explain: torch.Tensor, 
+                      feature_names: Optional[list[str]]=None, save_dir: Optional[Union[str, Path]] = None):
+    """
+    Calculates SHAP values and saves summary plots and data.
+
+    Args:
+        model (nn.Module): The trained PyTorch model.
+        background_data (torch.Tensor): A sample of data for the explainer background.
+        instances_to_explain (torch.Tensor): The specific data instances to explain.
+        feature_names (list of str | None): Names of the features for plot labeling.
+        save_dir (str | Path | None): Directory to save SHAP artifacts. If None, dot plot is shown.
+    """
+    print("\n--- SHAP Value Explanation ---")
+    print("Calculating SHAP values... ")
+    
+    model.eval()
+    model.cpu()
+    
+    explainer = shap.DeepExplainer(model, background_data)
+    shap_values = explainer.shap_values(instances_to_explain)
+
+    shap_values_for_plot = shap_values[1] if isinstance(shap_values, list) else shap_values
+    if isinstance(shap_values, list):
+        _LOGGER.info("Using SHAP values for the positive class (class 1) for plots.")
+
+    if save_dir:
+        save_dir_path = make_fullpath(save_dir, make=True)
+        # Save Bar Plot
+        bar_path = save_dir_path / "shap_bar_plot.svg"
+        shap.summary_plot(shap_values_for_plot, instances_to_explain, feature_names=feature_names, plot_type="bar", show=False)
+        plt.title("SHAP Feature Importance")
+        plt.tight_layout()
+        plt.savefig(bar_path)
+        _LOGGER.info(f"SHAP bar plot saved as '{bar_path.name}'")
+        plt.close()
+
+        # Save Dot Plot
+        dot_path = save_dir_path / "shap_dot_plot.svg"
+        shap.summary_plot(shap_values_for_plot, instances_to_explain, feature_names=feature_names, plot_type="dot", show=False)
+        plt.title("SHAP Feature Importance")
+        plt.tight_layout()
+        plt.savefig(dot_path)
+        _LOGGER.info(f"SHAP dot plot saved as '{dot_path.name}'")
+        plt.close()
+
+        # Save Summary Data to CSV
+        summary_path = save_dir_path / "shap_summary.csv"
+        mean_abs_shap = np.abs(shap_values_for_plot).mean(axis=0)
+        if feature_names is None:
+            feature_names = [f'feature_{i}' for i in range(len(mean_abs_shap))]
+        summary_df = pd.DataFrame({
+            'feature': feature_names,
+            'mean_abs_shap_value': mean_abs_shap
+        }).sort_values('mean_abs_shap_value', ascending=False)
+        summary_df.to_csv(summary_path, index=False)
+        _LOGGER.info(f"SHAP summary data saved as '{summary_path.name}'")
+    else:
+        _LOGGER.info("No save directory provided. Displaying SHAP dot plot.")
+        shap.summary_plot(shap_values_for_plot, instances_to_explain, feature_names=feature_names, plot_type="dot")