get_reprtree_from_rf_model Function Documentation

Your Name

2024-12-31

Purpose

The get_reprtree_from_rf_model function is designed to train a Random Forest model on a provided dataset and generate a representation tree (ReprTree) from the trained model. The function supports various configurations for data preprocessing, model hyperparameters, and sampling strategies, including random undersampling. Additionally, it allows for error correction and hyperparameter tuning.

Input Parameters

The following table describes the input parameters for the get_reprtree_from_rf_model function:

Parameter	Description	Type	Default Value
Data	A dataset to train the Random Forest model. If NULL, data is fetched using the get_Data_formatted_for_ml_and_best.m function.	Data frame	NULL
path_db	The path to the database used for fetching or processing the data.	Character string	-
rat_studies	A flag indicating whether rat studies are used.	Boolean (TRUE/FALSE)	FALSE
studyid_metadata	Metadata related to study IDs.	Data frame	NULL
fake_study	A flag indicating whether to use fake study data.	Boolean (TRUE/FALSE)	FALSE
use_xpt_file	A flag to indicate whether to use the XPT file format for data input.	Boolean (TRUE/FALSE)	FALSE
Round	A flag to round the data before processing.	Boolean (TRUE/FALSE)	FALSE
Impute	A flag to specify whether to impute missing values in the data.	Boolean (TRUE/FALSE)	FALSE
reps	The number of repetitions to perform for cross-validation or resampling.	Integer	-
holdback	The fraction of data to hold back for testing.	Numeric	-
Undersample	A flag indicating whether undersampling should be applied to balance the dataset.	Boolean (TRUE/FALSE)	FALSE
hyperparameter_tuning	A flag indicating whether hyperparameter tuning should be performed.	Boolean (TRUE/FALSE)	FALSE
error_correction_method	The method for error correction; valid options are 'Flip', 'Prune', or 'None'.	Character string	-

Output

The function generates a representation tree (ReprTree) from the trained Random Forest model and visualizes the first tree (k=5) from the model.

• A plot of the first tree from the Random Forest is displayed.
• The representation tree object is generated but not explicitly returned.

Key Steps

1. Data Preparation:
   • If the Data parameter is NULL, the function calls get_Data_formatted_for_ml_and_best.m to prepare the data for modeling.
   • Data is split into training and testing sets (70% for training and 30% for testing).
   • If undersampling is enabled (Undersample = TRUE), positive and negative samples are balanced in the training set by undersampling the majority class.
2. Model Training:
   • A Random Forest model is trained using the randomForest function. The target variable is Target_Organ, and the model uses the best hyperparameter (best.m) determined beforehand.
   • The number of trees in the forest is set to 500, and proximity calculations are enabled.
3. ReprTree Generation:
   • A ReprTree is generated using the reprtree::ReprTree function, which creates a representation of the trained Random Forest model.
   • The first tree (k=5) is plotted using reprtree::plot.getTree.
4. Visualization:
   • The first tree in the Random Forest model is visualized using the reprtree::plot.getTree function.

Example Usage

```r get_reprtree_from_rf_model( Data = my_data, path_db = “path/to/database”, rat_studies = TRUE, studyid_metadata = my_metadata, fake_study = FALSE, use_xpt_file = TRUE, Round = TRUE, Impute = TRUE, reps = 5, holdback = 0.3, Undersample = TRUE, hyperparameter_tuning = FALSE, error_correction_method = “Flip” )