comparison base_model_trainer.py @ 0:209b663a4f62 draft

planemo upload for repository https://github.com/goeckslab/gleam commit 5dd048419fcbd285a327f88267e93996cd279ee6

0:209b663a4f62

import base64
import logging
import os
import tempfile

import h5py
import joblib
import numpy as np
import pandas as pd
from feature_importance import FeatureImportanceAnalyzer
from sklearn.metrics import average_precision_score
from utils import get_html_closing, get_html_template

logging.basicConfig(level=logging.DEBUG)
LOG = logging.getLogger(__name__)


class BaseModelTrainer:

    def __init__(
            self,
            input_file,
            target_col,
            output_dir,
            task_type,
            random_seed,
            test_file=None,
            **kwargs):
        self.exp = None  # This will be set in the subclass
        self.input_file = input_file
        self.target_col = target_col
        self.output_dir = output_dir
        self.task_type = task_type
        self.random_seed = random_seed
        self.data = None
        self.target = None
        self.best_model = None
        self.results = None
        self.features_name = None
        self.plots = {}
        self.expaliner = None
        self.plots_explainer_html = None
        self.trees = []
        for key, value in kwargs.items():
            setattr(self, key, value)
        self.setup_params = {}
        self.test_file = test_file
        self.test_data = None

        LOG.info(f"Model kwargs: {self.__dict__}")

    def load_data(self):
        LOG.info(f"Loading data from {self.input_file}")
        self.data = pd.read_csv(self.input_file, sep=None, engine='python')
        self.data.columns = self.data.columns.str.replace('.', '_')

        numeric_cols = self.data.select_dtypes(include=['number']).columns
        non_numeric_cols = self.data.select_dtypes(exclude=['number']).columns

        self.data[numeric_cols] = self.data[numeric_cols].apply(
            pd.to_numeric, errors='coerce')

        if len(non_numeric_cols) > 0:
            LOG.info(f"Non-numeric columns found: {non_numeric_cols.tolist()}")

        names = self.data.columns.to_list()
        target_index = int(self.target_col) - 1
        self.target = names[target_index]
        self.features_name = [name
                              for i, name in enumerate(names)
                              if i != target_index]
        if hasattr(self, 'missing_value_strategy'):
            if self.missing_value_strategy == 'mean':
                self.data = self.data.fillna(
                    self.data.mean(numeric_only=True))
            elif self.missing_value_strategy == 'median':
                self.data = self.data.fillna(
                    self.data.median(numeric_only=True))
            elif self.missing_value_strategy == 'drop':
                self.data = self.data.dropna()
        else:
            # Default strategy if not specified
            self.data = self.data.fillna(self.data.median(numeric_only=True))

        if self.test_file:
            LOG.info(f"Loading test data from {self.test_file}")
            self.test_data = pd.read_csv(
                self.test_file, sep=None, engine='python')
            self.test_data = self.test_data[numeric_cols].apply(
                pd.to_numeric, errors='coerce')
            self.test_data.columns = self.test_data.columns.str.replace(
                '.', '_'
            )

    def setup_pycaret(self):
        LOG.info("Initializing PyCaret")
        self.setup_params = {
            'target': self.target,
            'session_id': self.random_seed,
            'html': True,
            'log_experiment': False,
            'system_log': False,
            'index': False,
        }

        if self.test_data is not None:
            self.setup_params['test_data'] = self.test_data

        if hasattr(self, 'train_size') and self.train_size is not None \
                and self.test_data is None:
            self.setup_params['train_size'] = self.train_size

        if hasattr(self, 'normalize') and self.normalize is not None:
            self.setup_params['normalize'] = self.normalize

        if hasattr(self, 'feature_selection') and \
                self.feature_selection is not None:
            self.setup_params['feature_selection'] = self.feature_selection

        if hasattr(self, 'cross_validation') and \
                self.cross_validation is not None \
                and self.cross_validation is False:
            self.setup_params['cross_validation'] = self.cross_validation

        if hasattr(self, 'cross_validation') and \
                self.cross_validation is not None:
            if hasattr(self, 'cross_validation_folds'):
                self.setup_params['fold'] = self.cross_validation_folds

        if hasattr(self, 'remove_outliers') and \
                self.remove_outliers is not None:
            self.setup_params['remove_outliers'] = self.remove_outliers

        if hasattr(self, 'remove_multicollinearity') and \
                self.remove_multicollinearity is not None:
            self.setup_params['remove_multicollinearity'] = \
                self.remove_multicollinearity

        if hasattr(self, 'polynomial_features') and \
                self.polynomial_features is not None:
            self.setup_params['polynomial_features'] = self.polynomial_features

        if hasattr(self, 'fix_imbalance') and \
                self.fix_imbalance is not None:
            self.setup_params['fix_imbalance'] = self.fix_imbalance

        LOG.info(self.setup_params)
        self.exp.setup(self.data, **self.setup_params)

    def train_model(self):
        LOG.info("Training and selecting the best model")
        if self.task_type == "classification":
            average_displayed = "Weighted"
            self.exp.add_metric(id=f'PR-AUC-{average_displayed}',
                                name=f'PR-AUC-{average_displayed}',
                                target='pred_proba',
                                score_func=average_precision_score,
                                average='weighted'
                                )

        if hasattr(self, 'models') and self.models is not None:
            self.best_model = self.exp.compare_models(
                include=self.models)
        else:
            self.best_model = self.exp.compare_models()
        self.results = self.exp.pull()
        if self.task_type == "classification":
            self.results.rename(columns={'AUC': 'ROC-AUC'}, inplace=True)

        _ = self.exp.predict_model(self.best_model)
        self.test_result_df = self.exp.pull()
        if self.task_type == "classification":
            self.test_result_df.rename(
                columns={'AUC': 'ROC-AUC'}, inplace=True)

    def save_model(self):
        hdf5_model_path = "pycaret_model.h5"
        with h5py.File(hdf5_model_path, 'w') as f:
            with tempfile.NamedTemporaryFile(delete=False) as temp_file:
                joblib.dump(self.best_model, temp_file.name)
                temp_file.seek(0)
                model_bytes = temp_file.read()
            f.create_dataset('model', data=np.void(model_bytes))

    def generate_plots(self):
        raise NotImplementedError("Subclasses should implement this method")

    def encode_image_to_base64(self, img_path):
        with open(img_path, 'rb') as img_file:
            return base64.b64encode(img_file.read()).decode('utf-8')

    def save_html_report(self):
        LOG.info("Saving HTML report")

        model_name = type(self.best_model).__name__
        excluded_params = ['html', 'log_experiment', 'system_log', 'test_data']
        filtered_setup_params = {
            k: v
            for k, v in self.setup_params.items() if k not in excluded_params
        }
        setup_params_table = pd.DataFrame(
            list(filtered_setup_params.items()), columns=['Parameter', 'Value']
        )

        best_model_params = pd.DataFrame(
            self.best_model.get_params().items(),
            columns=['Parameter', 'Value']
        )
        best_model_params.to_csv(
            os.path.join(self.output_dir, "best_model.csv"), index=False
        )
        self.results.to_csv(
            os.path.join(self.output_dir, "comparison_results.csv")
        )
        self.test_result_df.to_csv(
            os.path.join(self.output_dir, "test_results.csv")
        )

        plots_html = ""
        length = len(self.plots)
        for i, (plot_name, plot_path) in enumerate(self.plots.items()):
            encoded_image = self.encode_image_to_base64(plot_path)
            plots_html += f"""
            <div class="plot">
                <h3>{plot_name.capitalize()}</h3>
                <img src="data:image/png;base64,{encoded_image}"
                    alt="{plot_name}">
            </div>
            """
            if i < length - 1:
                plots_html += "<hr>"

        tree_plots = ""
        for i, tree in enumerate(self.trees):
            if tree:
                tree_plots += f"""
                <div class="plot">
                    <h3>Tree {i+1}</h3>
                    <img src="data:image/png;base64,
                    {tree}"
                    alt="tree {i+1}">
                </div>
                """

        analyzer = FeatureImportanceAnalyzer(
            data=self.data,
            target_col=self.target_col,
            task_type=self.task_type,
            output_dir=self.output_dir,
        )
        feature_importance_html = analyzer.run()

        html_content = f"""
        {get_html_template()}
            <h1>PyCaret Model Training Report</h1>
            <div class="tabs">
                <div class="tab" onclick="openTab(event, 'summary')">
                Setup & Best Model</div>
                <div class="tab" onclick="openTab(event, 'plots')">
                Best Model Plots</div>
                <div class="tab" onclick="openTab(event, 'feature')">
                Feature Importance</div>
        """
        if self.plots_explainer_html:
            html_content += """
                <div class="tab" onclick="openTab(event, 'explainer')">
                Explainer Plots</div>
            """
        html_content += f"""
            </div>
            <div id="summary" class="tab-content">
                <h2>Setup Parameters</h2>
                {setup_params_table.to_html(
                    index=False,
                    header=True,
                    classes='table sortable'
                )}
                <h5>If you want to know all the experiment setup parameters,
                  please check the PyCaret documentation for
                  the classification/regression <code>exp</code> function.</h5>
                <h2>Best Model: {model_name}</h2>
                {best_model_params.to_html(
                    index=False,
                    header=True,
                    classes='table sortable'
                )}
                <h2>Comparison Results on the Cross-Validation Set</h2>
                {self.results.to_html(index=False, classes='table sortable')}
                <h2>Results on the Test Set for the best model</h2>
                {self.test_result_df.to_html(
                    index=False,
                    classes='table sortable'
                )}
            </div>
            <div id="plots" class="tab-content">
                <h2>Best Model Plots on the testing set</h2>
                {plots_html}
            </div>
            <div id="feature" class="tab-content">
                {feature_importance_html}
            </div>
        """
        if self.plots_explainer_html:
            html_content += f"""
            <div id="explainer" class="tab-content">
                {self.plots_explainer_html}
                {tree_plots}
            </div>
            """
        html_content += """
        <script>
        document.addEventListener("DOMContentLoaded", function() {
            var tables = document.querySelectorAll("table.sortable");
            tables.forEach(function(table) {
                var headers = table.querySelectorAll("th");
                headers.forEach(function(header, index) {
                    header.style.cursor = "pointer";
                    // Add initial arrow (up) to indicate sortability
                    header.innerHTML += '<span class="sort-arrow"> ↑</span>';
                    header.addEventListener("click", function() {
                        var direction = this.getAttribute(
                            "data-sort-direction"
                        ) || "asc";
                        // Reset arrows in all headers of this table
                        headers.forEach(function(h) {
                            var arrow = h.querySelector(".sort-arrow");
                            if (arrow) arrow.textContent = " ↑";
                        });
                        // Set arrow for clicked header
                        var arrow = this.querySelector(".sort-arrow");
                        arrow.textContent = direction === "asc" ? " ↓" : " ↑";
                        sortTable(table, index, direction);
                        this.setAttribute("data-sort-direction",
                        direction === "asc" ? "desc" : "asc");
                    });
                });
            });
        });

        function sortTable(table, colNum, direction) {
            var tb = table.tBodies[0];
            var tr = Array.prototype.slice.call(tb.rows, 0);
            var multiplier = direction === "asc" ? 1 : -1;
            tr = tr.sort(function(a, b) {
                var aText = a.cells[colNum].textContent.trim();
                var bText = b.cells[colNum].textContent.trim();
                // Remove arrow from text comparison
                aText = aText.replace(/[↑↓]/g, '').trim();
                bText = bText.replace(/[↑↓]/g, '').trim();
                if (!isNaN(aText) && !isNaN(bText)) {
                    return multiplier * (
                        parseFloat(aText) - parseFloat(bText)
                    );
                } else {
                    return multiplier * aText.localeCompare(bText);
                }
            });
            for (var i = 0; i < tr.length; ++i) tb.appendChild(tr[i]);
        }
        </script>
        """
        html_content += f"""
        {get_html_closing()}
        """
        with open(
            os.path.join(self.output_dir, "comparison_result.html"),
            "w"
        ) as file:
            file.write(html_content)

    def save_dashboard(self):
        raise NotImplementedError("Subclasses should implement this method")

    def generate_plots_explainer(self):
        raise NotImplementedError("Subclasses should implement this method")

    # not working now
    def generate_tree_plots(self):
        from sklearn.ensemble import RandomForestClassifier, \
            RandomForestRegressor
        from xgboost import XGBClassifier, XGBRegressor
        from explainerdashboard.explainers import RandomForestExplainer

        LOG.info("Generating tree plots")
        X_test = self.exp.X_test_transformed.copy()
        y_test = self.exp.y_test_transformed

        is_rf = isinstance(self.best_model, RandomForestClassifier) or \
            isinstance(self.best_model, RandomForestRegressor)

        is_xgb = isinstance(self.best_model, XGBClassifier) or \
            isinstance(self.best_model, XGBRegressor)

        try:
            if is_rf:
                num_trees = self.best_model.n_estimators
            if is_xgb:
                num_trees = len(self.best_model.get_booster().get_dump())
            explainer = RandomForestExplainer(self.best_model, X_test, y_test)
            for i in range(num_trees):
                fig = explainer.decisiontree_encoded(tree_idx=i, index=0)
                LOG.info(f"Tree {i+1}")
                LOG.info(fig)
                self.trees.append(fig)
        except Exception as e:
            LOG.error(f"Error generating tree plots: {e}")

    def run(self):
        self.load_data()
        self.setup_pycaret()
        self.train_model()
        self.save_model()
        self.generate_plots()
        self.generate_plots_explainer()
        self.generate_tree_plots()
        self.save_html_report()
        # self.save_dashboard()