annotate tabular_learner.xml @ 8:ba45bc057d70 draft

planemo upload for repository https://github.com/goeckslab/gleam commit 1594d503179f28987720594eb49b48a15486f073

<tool id="tabular_learner" name="Tabular Learner" version="@TABULAR_LEARNER_VERSION@" profile="@PROFILE@">
    <description>applies and evaluates multiple machine learning models on a tabular dataset</description>
    <macros>
        <import>pycaret_macros.xml</import>
    </macros>
    <expand macro="python_requirements" />
    <command>
        <![CDATA[
        python $__tool_directory__/pycaret_train.py --input_file '$input_file' --target_col '$target_feature' --output_dir '.' --random_seed '$random_seed'
        #if $model_type == "classification"
            #if $classification_models
                --models '$classification_models'
            #end if
        #end if
        #if $model_type == "regression"
            #if $regression_models
                --models '$regression_models'
            #end if
        #end if
        #if $tune_model
            --tune_model
        #end if
        #if $customize_defaults == "true"
                #if $train_size
                --train_size '$train_size'
                #end if
                #if $normalize
                --normalize
                #end if
                #if $feature_selection
                --feature_selection
                #end if
                #if $enable_cross_validation == "true"
                    --cross_validation
                    #if $cross_validation_folds
                        --cross_validation_folds '$cross_validation_folds'
                    #end if
                #end if
                #if $enable_cross_validation == "false"
                    --no_cross_validation
                #end if
                #if $remove_outliers
                --remove_outliers
                #end if
                #if $remove_multicollinearity
                --remove_multicollinearity
                #end if
                #if $polynomial_features
                --polynomial_features
                #end if
                #if $fix_imbalance
                --fix_imbalance
                #end if
                #if $probability_threshold
                --probability_threshold '$probability_threshold'
                #end if
        #end if
        #if $has_test_file == "yes"
            --test_file '$test_file'
        #end if
        --model_type '$model_type'
        ]]>
    </command>
    <inputs>
        <param name="input_file" type="data" format="csv,tabular" label="Tabular Input Dataset" />
        <conditional name="test_data_choice">
            <param name="has_test_file" type="select" label="Do you have a separate test dataset?">
                <option value="no" selected="true">No</option>
                <option value="yes">Yes</option>
            </param>
            <when value="yes">
                <param name="test_file" type="data" format="csv,tabular" label="Tabular Test Dataset"
                       help="If a test dataset is provided, the input dataset will be split into training and validation sets only. 
                             If not, the tool will split your input into training, validation, and test automatically." />
            </when>
            <when value="no">
                <!-- Nothing extra shown -->
            </when>
        </conditional>
        <param name="target_feature" multiple="false" type="data_column" use_header_names="true" data_ref="input_file" label="Select the target column:" />
        <conditional name="model_selection">
            <param name="model_type" type="select" label="Task">
                <option value="classification">classification</option>
                <option value="regression">regression</option>
            </param>
            <when value="classification">
                <param name="classification_models" type="select" multiple="true" label="Only Select Classification Models if you don't want to compare all models">
                    <option value="lr">Logistic Regression</option>
                    <option value="knn">K Neighbors Classifier</option>
                    <option value="nb">Naive Bayes</option>
                    <option value="dt">Decision Tree Classifier</option>
                    <option value="svm">SVM - Linear Kernel</option>
                    <option value="rbfsvm">SVM - Radial Kernel</option>
                    <option value="gpc">Gaussian Process Classifier</option>
                    <option value="mlp">MLP Classifier</option>
                    <option value="ridge">Ridge Classifier</option>
                    <option value="rf">Random Forest Classifier</option>
                    <option value="qda">Quadratic Discriminant Analysis</option>
                    <option value="ada">Ada Boost Classifier</option>
                    <option value="gbc">Gradient Boosting Classifier</option>
                    <option value="lda">Linear Discriminant Analysis</option>
                    <option value="et">Extra Trees Classifier</option>
                    <option value="xgboost">Extreme Gradient Boosting</option>
                    <option value="lightgbm">Light Gradient Boosting Machine</option>
                    <option value="catboost">CatBoost Classifier</option>
                </param>
            </when>
            <when value="regression">
                <param name="regression_models" type="select" multiple="true" label="Only Select Regression Models if you don't want to compare all models">
                    <option value="lr">Linear Regression</option>
                    <option value="lasso">Lasso Regression</option>
                    <option value="ridge">Ridge Regression</option>
                    <option value="en">Elastic Net</option>
                    <option value="lar">Least Angle Regression</option>
                    <option value="llar">Lasso Least Angle Regression</option>
                    <option value="omp">Orthogonal Matching Pursuit</option>
                    <option value="br">Bayesian Ridge</option>
                    <option value="ard">Automatic Relevance Determination</option>
                    <option value="par">Passive Aggressive Regressor</option>
                    <option value="ransac">Random Sample Consensus</option>
                    <option value="tr">TheilSen Regressor</option>
                    <option value="huber">Huber Regressor</option>
                    <option value="kr">Kernel Ridge</option>
                    <option value="svm">Support Vector Regression</option>
                    <option value="knn">K Neighbors Regressor</option>
                    <option value="dt">Decision Tree Regressor</option>
                    <option value="rf">Random Forest Regressor</option>
                    <option value="et">Extra Trees Regressor</option>
                    <option value="ada">AdaBoost Regressor</option>
                    <option value="gbr">Gradient Boosting Regressor</option>
                    <option value="mlp">MLP Regressor</option>
                    <option value="xgboost">Extreme Gradient Boosting</option>
                    <option value="lightgbm">Light Gradient Boosting Machine</option>
                    <option value="catboost">CatBoost Regressor</option>
                </param>
            </when>
        </conditional>
        <param name="tune_model" type="boolean" truevalue="True" falsevalue="False" label="Tune hyperparameters" help="Hyperparameter tuning on the best model" />
        <param name="random_seed" type="integer" value="42" label="Random Seed" help="Random seed for reproducibility." />
        <conditional name="advanced_settings">
            <param name="customize_defaults" type="select" label="Customize Default Settings?" help="Select yes if you want to customize the default settings of the experiment.">
                <option value="false" selected="true">No</option>
                <option value="true">Yes</option>
            </param>
            <when value="true">
                <param name="train_size" type="float" value="0.7" min="0.1" max="0.9" label="Train Size" help="Proportion of the input dataset to include in the train split." />
                <param name="normalize" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Normalize Data" help="Whether to normalize data before training." />
                <param name="feature_selection" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Feature Selection" help="Whether to perform feature selection." />
                <conditional name="cross_validation">
                    <param name="enable_cross_validation" type="select" label="Enable Cross Validation?" help="Select whether to enable cross-validation." >
                        <option value="false" >No</option>
                        <option value="true" selected="true">Yes</option>
                    </param>
                    <when value="true">
                        <param name="cross_validation_folds" type="integer" value="10" min="2" max="20" label="Cross Validation Folds" help="Number of folds to use for cross-validation." />
                    </when>
                    <when value="false">
                        <!-- No additional parameters to show if the user selects 'No' -->
                    </when>
                </conditional>
                <param name="remove_outliers" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Remove Outliers" help="Whether to remove outliers from the input dataset before training." />
                <param name="remove_multicollinearity" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Remove Multicollinearity" help="Whether to remove multicollinear features before training." />
                <param name="polynomial_features" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Polynomial Features" help="Whether to create polynomial features before training." />
                <param name="fix_imbalance" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Fix Imbalance" help="ONLY for classfication! Whether to use SMOTE or similar methods to fix imbalance in the input dataset." />
                <param name="probability_threshold" type="float" min="0.0" max="1.0" value="0.5" label="Classification Probability Threshold" help="Only applies to classification. Probability above which a prediction is considered positive. Default is 0.5." />
            </when>
            <when value="false">
                <!-- No additional parameters to show if the user selects 'No' -->
            </when>
        </conditional>
    </inputs>
    <outputs>
        <data name="model" format="h5" from_work_dir="pycaret_model.h5" label="${tool.name} best model on ${on_string}" />
        <data name="best_model_csv" format="csv" from_work_dir="best_model.csv" label="${tool.name} The parameters of the best model on ${on_string}" hidden="true" />
        <data name="comparison_result" format="html" from_work_dir="comparison_result.html" label="${tool.name} analysis report on ${on_string}" />
    </outputs>
    <tests>
        <test>
            <param name="input_file" value="pcr.tsv"/>
            <param name="target_feature" value="11"/> 
            <param name="model_type" value="classification"/>
            <param name="random_seed" value="42"/>
            <param name="customize_defaults" value="true"/>
            <param name="train_size" value="0.8"/>
            <param name="normalize" value="true"/>
            <param name="feature_selection" value="true"/>
            <param name="enable_cross_validation" value="true"/>
            <param name="cross_validation_folds" value="5"/>
            <param name="remove_outliers" value="true"/>
            <param name="remove_multicollinearity" value="true"/>
            <param name="probability_threshold" value="0.4" />
            <output name="model" file="expected_model_classification_customized.h5" compare="sim_size"/>
            <output name="comparison_result">
                <assert_contents>
                    <has_text text="Validation Summary" />
                    <has_text text="Test Summary" />
                    <has_text text="Feature Importance" />
                </assert_contents>
            </output>
            <output name="best_model_csv" value="expected_best_model_classification_customized.csv" />
        </test>
        <test>
            <param name="input_file" value="pcr.tsv"/>
            <param name="target_feature" value="11"/> 
            <param name="model_type" value="classification"/>
            <param name="random_seed" value="42"/>
            <param name="customize_defaults" value="true"/>
            <param name="train_size" value="0.8"/>
            <param name="normalize" value="true"/>
            <param name="feature_selection" value="true"/>
            <param name="enable_cross_validation" value="false"/>
            <param name="remove_outliers" value="true"/>
            <param name="remove_multicollinearity" value="true"/>
            <param name="probability_threshold" value="0.6" />
            <output name="model" file="expected_model_classification_customized_cross_off.h5" compare="sim_size"/>
            <output name="comparison_result">
                <assert_contents>
                    <has_text text="Validation Summary" />
                    <has_text text="Test Summary" />
                    <has_text text="Feature Importance" />
                </assert_contents>
            </output>
            <output name="best_model_csv" value="expected_best_model_classification_customized_cross_off.csv" />
        </test>
        <test>
            <param name="input_file" value="pcr.tsv"/>
            <param name="target_feature" value="11"/> 
            <param name="model_type" value="classification"/>
            <param name="random_seed" value="42"/>
            <param name="tune_model" value="true"/>
            <output name="model" file="expected_model_classification.h5" compare="sim_size"/>
            <output name="comparison_result"> 
                <assert_contents>
                    <has_text text="Validation Summary" />
                    <has_text text="Test Summary" />
                    <has_text text="Feature Importance" />
                </assert_contents>
            </output>
            <output name="best_model_csv" value="expected_best_model_classification.csv" />
        </test>
        <test>
            <param name="input_file" value="pcr.tsv"/>
            <param name="target_feature" value="11"/> 
            <param name="model_type" value="classification"/>
            <param name="random_seed" value="42"/>
            <output name="model" file="expected_model_classification.h5" compare="sim_size"/>
            <output name="comparison_result"> 
                <assert_contents>
                    <has_text text="Validation Summary" />
                    <has_text text="Test Summary" />
                    <has_text text="Feature Importance" />
                </assert_contents>
            </output>
            <output name="best_model_csv" value="expected_best_model_classification.csv" />
        </test>
        <test>
            <param name="input_file" value="auto-mpg.tsv"/>
            <param name="target_feature" value="1"/> 
            <param name="model_type" value="regression"/>
            <param name="random_seed" value="42"/>
            <output name="model" file="expected_model_regression.h5" compare="sim_size" />
            <output name="comparison_result">
                <assert_contents>
                    <has_text text="Validation Summary" />
                    <has_text text="Test Summary" />
                    <has_text text="Feature Importance" />
                </assert_contents>
            </output>
            <output name="best_model_csv" value="expected_best_model_regression.csv" />
        </test>
    </tests>
    <help>
        This tool uses PyCaret to train and evaluate machine learning models.
        It compares different models on a dataset and provides the best model based on the performance metrics.

        **Outputs**

        - **Model**: The best model trained on the dataset in h5 format.


        - **Comparison Result**: The comparison result of different models in html format. 
            It contains the performance metrics of different models, plots of the best model 
            on the testing set (or part of the training set if a separate test set is not uploaded), and feature analysis plots.

    </help>
    <expand macro="macro_citations" />
</tool>