EDA Toolkit 0.0.10

Version 0.0.10

Legend Handling:

• The legend is now displayed only if there are valid legend handles (len(handles) > 0) and if show_legend is set to True.
• The check ax.get_legend().remove() ensures that unnecessary legends are removed if they are empty or if show_legend is set to False.

Error Handling:

• Error handling in the except block has been enhanced to ensure that any exceptions related to legends or labels are managed properly. The legend handling logic still respects the show_legend flag even in cases where exceptions occur.

This update prevents empty legend squares from appearing and maintains the intended default behavior of showing legends only when they contain relevant content.

EDA Toolkit 0.0.9

Version 0.0.9

Bug Fixes and Minor Improvements

Improved error messages and validation checks across multiple functions to prevent common pitfalls and ensure smoother user experience.

Visualization Enhancements

DataFrame Columns: Added a background_color variable to dataframe_columns, allowing the user to enter a string representing a color name, or hex value. Try/Except on the output, in case the end user has a deprecated version of Pandas, where the styler would use hide() instead of hide_index(). The highlighted columns allow for easier null versus unique value analysis.

The docstring now clearly describes the purpose of the function—analyzing DataFrame columns to provide summary statistics.

Args:

• The df argument is specified as a pandas.DataFrame.
• The background_color argument is marked as optional, with a brief description of its role.
• The return_df argument is also marked as optional, explaining what it controls.

Returns: The return type is specified as pandas.DataFrame, with a clear explanation of the difference based on the return_df flag.

KDE Distribution Plots: Improved kde_distributions() with enhanced options for log scaling, mean/median plotting, custom standard deviation lines, and better handling of legends and scientific notation.

Scatter Plots: Enhanced scatter_fit_plot() with support for hue-based coloring, best fit lines, correlation display, and flexible grid plotting options.

EDA Toolkit 0.0.8e

Version 0.0.8e

This update introduces several key changes to the plot_3d_pdp function and minor changes to the stacked_crosstab_plot function, simplifying the function's interface and improving usability, while maintaining the flexibility needed for diverse visualization needs.

stacked_crosstab_plot

• Flexible save_formats Input:

  • save_formats now accepts a string, tuple, or list for specifying formats (e.g., "png", ("png", "svg"), or ["png", "svg"]).
  • Single strings or tuples are automatically converted to lists for consistent processing.

• Dynamic Error Handling:

  • Added checks to ensure a valid path is provided for each format in save_formats.
  • Raises a ValueError if a format is specified without a corresponding path, with a clear, dynamic error message.

• Improved Plot Saving Logic:

  • Updated logic allows saving plots in one format (e.g., only "png" or "svg") without requiring the other.
  • Simplified and more intuitive path handling for saving plots.

plot_3d_pdp

1. Parameter Changes

---

• Removed Parameters:

  • The parameters x_label_plotly, y_label_plotly, and z_label_plotly have been removed. These parameters previously allowed custom axis labels specifically for the Plotly plot, defaulting to the general x_label, y_label, and z_label. Removing these parameters simplifies the function signature while maintaining flexibility.

• Default Values for Labels:

  • The parameters x_label, y_label, and z_label are now optional, with None as the default. If not provided, these labels will automatically default to the names of the features in the feature_names_list. This change makes the function more user-friendly, particularly for cases where default labels are sufficient.

• Changes in Default Values for View Angles:

  • The default values for camera positioning parameters have been updated: horizontal is now -1.25, depth is now 1.25, and vertical is now 1.25. These adjustments refine the default 3D view perspective for the Plotly plot, providing a more intuitive starting view.

2. Plot Generation Logic

---

• Conditionally Checking Labels:

  • The function now checks whether x_label, y_label, and z_label are provided. If these are None, the function will automatically assign default labels based on the feature_names_list. This enhancement reduces the need for users to manually specify labels, making the function more adaptive.

• Camera Position Adjustments:

  • The camera positions for the Plotly plot are now adjusted by multiplying horizontal, depth, and vertical by zoom_out_factor. This change allows for more granular control over the 3D view, enhancing the interactivity and flexibility of the Plotly visualizations.

• Surface Plot Coordinates Adjustments:

  • The order of the coordinates for the Plotly plot’s surface has been changed from ZZ, XX, YY[::-1] to ZZ, XX, YY. This adjustment ensures the proper alignment of axes and grids, resulting in more accurate visual representations.

3. Code Simplifications

---

• Removed Complexity:

  • By removing the x_label_plotly, y_label_plotly, and z_label_plotly parameters, the code is now simpler and easier to maintain. This change reduces potential confusion and streamlines the function for users who do not need distinct labels for Matplotlib and Plotly plots.

• Fallback Mechanism for Grid Values:

  • The function continues to implement a fallback mechanism when extracting grid values, ensuring compatibility with various versions of scikit-learn. This makes the function robust across different environments.

4. Style Adjustments

---

• Label Formatting:

  • The new version consistently uses y_label, x_label, and z_label for axis labels in the Matplotlib plot, aligning the formatting across different plot types.

• Color Bar Adjustments:

  • The color bar configuration in the Matplotlib plot has been slightly adjusted with a shrink value of 0.6 and a pad value of 0.02. These adjustments result in a more refined visual appearance, particularly in cases where space is limited.

5. Potential Use Case Differences

---

• Simplified Interface:

  • The updated function is more streamlined for users who prefer a simplified interface without the need for separate label customizations for Plotly and Matplotlib plots. This makes it easier to use in common scenarios.

• Less Granular Control:

  • Users who need more granular control, particularly for presentations or specific formatting, may find the older version more suitable. The removal of the *_plotly label parameters means that all plots now use the same labels across Matplotlib and Plotly.

6. Matplotlib Plot Adjustments

---

• Wireframe and Surface Plot Enhancements:

  • The logic for plotting wireframes and surface plots in Matplotlib remains consistent with previous versions, with subtle enhancements to color and layout management to improve overall aesthetics.

Summary

• Version 0.0.8e of the plot_3d_pdp function introduces simplifications that reduce the number of parameters and streamline the plotting process. While some customizability has been removed, the function remains flexible enough for most use cases and is easier to use.
• Key updates include adjusted default camera views for 3D plots, removal of Plotly-specific label parameters, and improved automatic labeling and plotting logic.

EDA Toolkit 0.0.8

Version 0.0.8

This update introduces several key changes to the plot_3d_pdp function and minor changes to the stacked_crosstab_plot function, simplifying the function's interface and improving usability, while maintaining the flexibility needed for diverse visualization needs.

stacked_crosstab_plot

• Flexible save_formats Input:

  • save_formats now accepts a string, tuple, or list for specifying formats (e.g., "png", ("png", "svg"), or ["png", "svg"]).
  • Single strings or tuples are automatically converted to lists for consistent processing.

• Dynamic Error Handling:

  • Added checks to ensure a valid path is provided for each format in save_formats.
  • Raises a ValueError if a format is specified without a corresponding path, with a clear, dynamic error message.

• Improved Plot Saving Logic:

  • Updated logic allows saving plots in one format (e.g., only "png" or "svg") without requiring the other.
  • Simplified and more intuitive path handling for saving plots.

plot_3d_pdp

1. Parameter Changes

---

• Removed Parameters:

  • The parameters x_label_plotly, y_label_plotly, and z_label_plotly have been removed. These parameters previously allowed custom axis labels specifically for the Plotly plot, defaulting to the general x_label, y_label, and z_label. Removing these parameters simplifies the function signature while maintaining flexibility.

• Default Values for Labels:

  • The parameters x_label, y_label, and z_label are now optional, with None as the default. If not provided, these labels will automatically default to the names of the features in the feature_names_list. This change makes the function more user-friendly, particularly for cases where default labels are sufficient.

• Changes in Default Values for View Angles:

  • The default values for camera positioning parameters have been updated: horizontal is now -1.25, depth is now 1.25, and vertical is now 1.25. These adjustments refine the default 3D view perspective for the Plotly plot, providing a more intuitive starting view.

2. Plot Generation Logic

---

• Conditionally Checking Labels:

  • The function now checks whether x_label, y_label, and z_label are provided. If these are None, the function will automatically assign default labels based on the feature_names_list. This enhancement reduces the need for users to manually specify labels, making the function more adaptive.

• Camera Position Adjustments:

  • The camera positions for the Plotly plot are now adjusted by multiplying horizontal, depth, and vertical by zoom_out_factor. This change allows for more granular control over the 3D view, enhancing the interactivity and flexibility of the Plotly visualizations.

• Surface Plot Coordinates Adjustments:

  • The order of the coordinates for the Plotly plot’s surface has been changed from ZZ, XX, YY[::-1] to ZZ, XX, YY. This adjustment ensures the proper alignment of axes and grids, resulting in more accurate visual representations.

3. Code Simplifications

---

• Removed Complexity:

  • By removing the x_label_plotly, y_label_plotly, and z_label_plotly parameters, the code is now simpler and easier to maintain. This change reduces potential confusion and streamlines the function for users who do not need distinct labels for Matplotlib and Plotly plots.

• Fallback Mechanism for Grid Values:

  • The function continues to implement a fallback mechanism when extracting grid values, ensuring compatibility with various versions of scikit-learn. This makes the function robust across different environments.

4. Style Adjustments

---

• Label Formatting:

  • The new version consistently uses y_label, x_label, and z_label for axis labels in the Matplotlib plot, aligning the formatting across different plot types.

• Color Bar Adjustments:

  • The color bar configuration in the Matplotlib plot has been slightly adjusted with a shrink value of 0.6 and a pad value of 0.02. These adjustments result in a more refined visual appearance, particularly in cases where space is limited.

5. Potential Use Case Differences

---

• Simplified Interface:

  • The updated function is more streamlined for users who prefer a simplified interface without the need for separate label customizations for Plotly and Matplotlib plots. This makes it easier to use in common scenarios.

• Less Granular Control:

  • Users who need more granular control, particularly for presentations or specific formatting, may find the older version more suitable. The removal of the *_plotly label parameters means that all plots now use the same labels across Matplotlib and Plotly.

6. Matplotlib Plot Adjustments

---

• Wireframe and Surface Plot Enhancements:

  • The logic for plotting wireframes and surface plots in Matplotlib remains consistent with previous versions, with subtle enhancements to color and layout management to improve overall aesthetics.

Summary

• Version 0.0.8 of the plot_3d_pdp function introduces simplifications that reduce the number of parameters and streamline the plotting process. While some customizability has been removed, the function remains flexible enough for most use cases and is easier to use.
• Key updates include adjusted default camera views for 3D plots, removal of Plotly-specific label parameters, and improved automatic labeling and plotting logic.

EDA Toolkit 0.0.8d

Version 0.0.8d

This update introduces several key changes to the plot_3d_pdp function and minor changes to the stacked_crosstab_plot function, simplifying the function's interface and improving usability, while maintaining the flexibility needed for diverse visualization needs.

stacked_crosstab_plot

• Flexible save_formats Input:

  • save_formats now accepts a string, tuple, or list for specifying formats (e.g., "png", ("png", "svg"), or ["png", "svg"]).
  • Single strings or tuples are automatically converted to lists for consistent processing.

• Dynamic Error Handling:

  • Added checks to ensure a valid path is provided for each format in save_formats.
  • Raises a ValueError if a format is specified without a corresponding path, with a clear, dynamic error message.

• Improved Plot Saving Logic:

  • Updated logic allows saving plots in one format (e.g., only "png" or "svg") without requiring the other.
  • Simplified and more intuitive path handling for saving plots.

plot_3d_pdp

1. Parameter Changes

---

• Removed Parameters:

  • The parameters x_label_plotly, y_label_plotly, and z_label_plotly have been removed. These parameters previously allowed custom axis labels specifically for the Plotly plot, defaulting to the general x_label, y_label, and z_label. Removing these parameters simplifies the function signature while maintaining flexibility.

• Default Values for Labels:

  • The parameters x_label, y_label, and z_label are now optional, with None as the default. If not provided, these labels will automatically default to the names of the features in the feature_names_list. This change makes the function more user-friendly, particularly for cases where default labels are sufficient.

• Changes in Default Values for View Angles:

  • The default values for camera positioning parameters have been updated: horizontal is now -1.25, depth is now 1.25, and vertical is now 1.25. These adjustments refine the default 3D view perspective for the Plotly plot, providing a more intuitive starting view.

2. Plot Generation Logic

---

• Conditionally Checking Labels:

  • The function now checks whether x_label, y_label, and z_label are provided. If these are None, the function will automatically assign default labels based on the feature_names_list. This enhancement reduces the need for users to manually specify labels, making the function more adaptive.

• Camera Position Adjustments:

  • The camera positions for the Plotly plot are now adjusted by multiplying horizontal, depth, and vertical by zoom_out_factor. This change allows for more granular control over the 3D view, enhancing the interactivity and flexibility of the Plotly visualizations.

• Surface Plot Coordinates Adjustments:

  • The order of the coordinates for the Plotly plot’s surface has been changed from ZZ, XX, YY[::-1] to ZZ, XX, YY. This adjustment ensures the proper alignment of axes and grids, resulting in more accurate visual representations.

3. Code Simplifications

---

• Removed Complexity:

  • By removing the x_label_plotly, y_label_plotly, and z_label_plotly parameters, the code is now simpler and easier to maintain. This change reduces potential confusion and streamlines the function for users who do not need distinct labels for Matplotlib and Plotly plots.

• Fallback Mechanism for Grid Values:

  • The function continues to implement a fallback mechanism when extracting grid values, ensuring compatibility with various versions of scikit-learn. This makes the function robust across different environments.

4. Style Adjustments

---

• Label Formatting:

  • The new version consistently uses y_label, x_label, and z_label for axis labels in the Matplotlib plot, aligning the formatting across different plot types.

• Color Bar Adjustments:

  • The color bar configuration in the Matplotlib plot has been slightly adjusted with a shrink value of 0.6 and a pad value of 0.02. These adjustments result in a more refined visual appearance, particularly in cases where space is limited.

5. Potential Use Case Differences

---

• Simplified Interface:

  • The updated function is more streamlined for users who prefer a simplified interface without the need for separate label customizations for Plotly and Matplotlib plots. This makes it easier to use in common scenarios.

• Less Granular Control:

  • Users who need more granular control, particularly for presentations or specific formatting, may find the older version more suitable. The removal of the *_plotly label parameters means that all plots now use the same labels across Matplotlib and Plotly.

6. Matplotlib Plot Adjustments

---

• Wireframe and Surface Plot Enhancements:

  • The logic for plotting wireframes and surface plots in Matplotlib remains consistent with previous versions, with subtle enhancements to color and layout management to improve overall aesthetics.

Summary

• Version 0.0.8d of the plot_3d_pdp function introduces simplifications that reduce the number of parameters and streamline the plotting process. While some customizability has been removed, the function remains flexible enough for most use cases and is easier to use.
• Key updates include adjusted default camera views for 3D plots, removal of Plotly-specific label parameters, and improved automatic labeling and plotting logic.

EDA Toolkit 0.0.8c

EDA Toolkit 0.0.8c:

Summary of Changes:

1. New Features & Enhancements:

• plot_3d_pdp Function:

  • Added show_modebar Parameter: Introduced a new boolean parameter, show_modebar, to allow users to toggle the visibility of the mode bar in Plotly interactive plots.
  • Custom Margins and Layout Adjustments:
    • Added parameters for left_margin, right_margin, and top_margin to provide users with more control over the plot layout in Plotly.
    • Adjusted default values and added options for better customization of the Plotly color bar (cbar_x, cbar_thickness) and title positioning (title_x, title_y).
  • Plotly Configuration:
    • Enhanced the configuration options to allow users to enable or disable zoom functionality (enable_zoom) in the interactive Plotly plots.
    • Updated the code to reflect these new parameters, allowing for greater flexibility in the appearance and interaction with the Plotly plots.
  • Error Handling:
    • Added input validation for html_file_path and html_file_name to ensure these are provided when necessary based on the selected plot_type.

• plot_2d_pdp Function:

  • Introduced file_prefix Parameter:
    • Added a new file_prefix parameter to allow users to specify a prefix for filenames when saving grid plots. This change streamlines the naming process for saved plots and improves file organization.
  • Enhanced Plot Type Flexibility:
    • The plot_type parameter now includes an option to generate both grid and individual plots (both). This feature allows users to create a combination of both layout styles in one function call.
    • Updated input validation and logic to handle this new option effectively.
  • Added save_plots Parameter:
    • Introduced a new parameter, save_plots, to control the saving of plots. Users can specify whether to save all plots, only individual plots, only grid plots, or none.
  • Custom Margins and Layout Adjustments:
    • Included the save_plots parameter in the validation process to ensure paths are provided when needed for saving the plots.

2. Documentation Updates:

• Docstrings:
  • Updated docstrings for both functions to reflect the new parameters and enhancements, providing clearer and more comprehensive guidance for users.
  • Detailed the use of new parameters such as show_modebar, file_prefix, save_plots, and others, ensuring that the function documentation is up-to-date with the latest changes.

3. Refactoring & Code Cleanup:

• Code Structure:
  • Improved the code structure to maintain clarity and readability, particularly around the new functionality.
  • Consolidated the layout configuration settings for the Plotly plots into a more flexible and user-friendly format, making it easier for users to customize their plots.

---

This version enhances the usability of the plot_3d_pdp and plot_2d_pdp functions, introduces new features for greater flexibility in plot customization, and ensures that the functions are well-documented and easy to use. The updates are backward-compatible and aim to provide a more seamless user experience in generating and saving both 3D and 2D partial dependence plots.

EDA Toolkit 0.0.8b

Version 0.0.8b Release Notes

We are excited to announce the release of version 0.0.8b, which introduces significant enhancements and new features to improve the usability and functionality of our toolkit.

New Features:

1. Optional file_prefix in stacked_crosstab_plot Function

   • The stacked_crosstab_plot function has been updated to make the file_prefix argument optional. If the user does not provide a file_prefix, the function will now automatically generate a default prefix based on the col and func_col parameters. This change streamlines the process of generating plots by reducing the number of required arguments.
   • Key Improvement:
     • Users can now omit the file_prefix argument, and the function will still produce appropriately named plot files, enhancing ease of use.
     • Backward compatibility is maintained, allowing users who prefer to specify a custom file_prefix to continue doing so without any issues.

2. Introduction of 3D and 2D Partial Dependence Plot Functions

   • Two new functions, plot_3d_pdp and plot_2d_pdp, have been added to the toolkit, expanding the visualization capabilities for machine learning models.
     • plot_3d_pdp: Generates 3D partial dependence plots for two features, supporting both static visualizations (using Matplotlib) and interactive plots (using Plotly). The function offers extensive customization options, including labels, color maps, and saving formats.
     • plot_2d_pdp: Creates 2D partial dependence plots for specified features with flexible layout options (grid or individual plots) and customization of figure size, font size, and saving formats.
   • Key Features:
     • Compatibility: Both functions are compatible with various versions of scikit-learn, ensuring broad usability.
     • Customization: Extensive options for customizing visual elements, including figure size, font size, and color maps.
     • Interactive 3D Plots: The plot_3d_pdp function supports interactive visualizations, providing an enhanced user experience for exploring model predictions in 3D space.

Impact:

• These updates improve the user experience by reducing the complexity of function calls and introducing powerful new tools for model interpretation.
• The optional file_prefix enhancement simplifies plot generation while maintaining the flexibility to define custom filenames.
• The new partial dependence plot functions offer robust visualization options, making it easier to analyze and interpret the influence of specific features in machine learning models.

We encourage users to explore these new features and provide feedback on their experience. As always, we remain committed to continuous improvement and welcome suggestions for future updates.

EDA Toolkit 0.0.8a

Version 0.0.8 Release Notes

We are excited to announce the release of version 0.0.8, which introduces significant enhancements and new features to improve the usability and functionality of our toolkit.

New Features:

1. Optional file_prefix in stacked_crosstab_plot Function

   • The stacked_crosstab_plot function has been updated to make the file_prefix argument optional. If the user does not provide a file_prefix, the function will now automatically generate a default prefix based on the col and func_col parameters. This change streamlines the process of generating plots by reducing the number of required arguments.
   • Key Improvement:
     • Users can now omit the file_prefix argument, and the function will still produce appropriately named plot files, enhancing ease of use.
     • Backward compatibility is maintained, allowing users who prefer to specify a custom file_prefix to continue doing so without any issues.

2. Introduction of 3D and 2D Partial Dependence Plot Functions

   • Two new functions, plot_3d_pdp and plot_2d_pdp, have been added to the toolkit, expanding the visualization capabilities for machine learning models.
     • plot_3d_pdp: Generates 3D partial dependence plots for two features, supporting both static visualizations (using Matplotlib) and interactive plots (using Plotly). The function offers extensive customization options, including labels, color maps, and saving formats.
     • plot_2d_pdp: Creates 2D partial dependence plots for specified features with flexible layout options (grid or individual plots) and customization of figure size, font size, and saving formats.
   • Key Features:
     • Compatibility: Both functions are compatible with various versions of scikit-learn, ensuring broad usability.
     • Customization: Extensive options for customizing visual elements, including figure size, font size, and color maps.
     • Interactive 3D Plots: The plot_3d_pdp function supports interactive visualizations, providing an enhanced user experience for exploring model predictions in 3D space.

Impact:

• These updates improve the user experience by reducing the complexity of function calls and introducing powerful new tools for model interpretation.
• The optional file_prefix enhancement simplifies plot generation while maintaining the flexibility to define custom filenames.
• The new partial dependence plot functions offer robust visualization options, making it easier to analyze and interpret the influence of specific features in machine learning models.

We encourage users to explore these new features and provide feedback on their experience. As always, we remain committed to continuous improvement and welcome suggestions for future updates.

EDA Toolkit 0.0.7

Add flex_corr_matrix function for customizable correlation matrix visualization

This release introduces a new function, flex_corr_matrix, which allows users to generate both full and upper triangular correlation heatmaps with a high degree of customization. The function includes options to annotate the heatmap, save the plots, and pass additional parameters to seaborn.heatmap().

Summary of Changes:

• New Function: flex_corr_matrix
  • Functionality:
    • Generates a correlation heatmap for a given DataFrame.
    • Supports both full and upper triangular correlation matrices based on the triangular parameter.
    • Allows users to customize various aspects of the plot, including colormap, figure size, axis label rotation, and more.
    • Accepts additional keyword arguments via **kwargs to pass directly to seaborn.heatmap().
    • Includes validation to ensure the triangular, annot, and save_plots parameters are boolean values.
    • Raises an exception if save_plots=True but neither image_path_png nor image_path_svg is specified.

Usage:

# Full correlation matrix example
flex_corr_matrix(df=my_dataframe, triangular=False, cmap="coolwarm", annot=True)

# Upper triangular correlation matrix example
flex_corr_matrix(df=my_dataframe, triangular=True, cmap="coolwarm", annot=True)

Contingency table df to object type

Convert all columns in dataframe to object, to prevent issues with numerical columns.

df = df.astype(str).fillna("")

EDA Toolkit 0.0.6

Add validation for plot_type parameter in kde_distributions function

This release adds a validation step for the plot_type parameter in the kde_distributions function. The allowed values for plot_type are "hist", "kde", and "both". If an invalid value is provided, the function will now raise a ValueError with a clear message indicating the accepted values. This change improves the robustness of the function and helps prevent potential errors due to incorrect parameter values.

    # Validate plot_type parameter
    valid_plot_types = ["hist", "kde", "both"]
    if plot_type.lower() not in valid_plot_types:
        raise ValueError(
            f"Invalid plot_type value. Expected one of {valid_plot_types}, "
            f"got '{plot_type}' instead."
        )