Configuration File Reference

The assets/config.yml file is the central place to configure the ConversionFlow pipeline. This document provides a reference for all the parameters in this file.

project_name

  • Type: string

  • Description: The name of your project. Used for titling outputs and reports.

preprocess

This section controls the preprocessing of the data.

  • scaling:

    • enabled: boolean - Whether to enable scaling of numerical features.

    • method: string - The scaling method to use (e.g., standard).

  • missing_values:

    • strategy: string - The strategy for handling missing values (e.g., fill).

    • fill_value: any - The value to use when filling missing values.

  • exclude_from_scaling: list - A list of column names to exclude from scaling.

api

  • gemini_key: string - Your API key for the Google Gemini API, used for automated analysis.

model

This section defines the structure of the Bayesian Network.

  • name: string - A unique name for your model.

  • description: string - A brief description of the model.

  • nodes: dict - A dictionary where keys are stage names and values are lists of node names.

  • edges: list - A list of lists, where each inner list represents a directed edge from a parent to a child node.

priors

This section defines the prior distributions for the model parameters.

  • beta_distributions: dict - A nested dictionary defining the priors for the beta coefficients.

    • The top-level key is the parent node.

    • The nested key is the child node.

    • The value is a dictionary specifying the distribution (e.g., Normal, HalfCauchy) and its parameters (e.g., mu, sigma).

inference

This section controls the MCMC inference process.

  • draws: int - The number of samples to draw from the posterior distribution.

  • tune: int - The number of tuning (burn-in) steps to discard.

  • chains: int - The number of independent Markov chains to run.

  • random_seed: int - A random seed for reproducibility.

  • target_accept: float - The target acceptance rate for the NUTS sampler.

  • max_treedepth: int - The maximum tree depth for the NUTS sampler.

output

This section controls the output of the pipeline.

  • directory: string - The directory where the output files will be saved.

  • figures:

    • extension: string - The file extension for the output plots (e.g., .png).

    • dpi: int - The resolution of the output plots.

logging

This section controls the logging configuration.

  • level: string - The logging level (e.g., INFO, DEBUG).

  • file: string - The path to the log file.