cyclic_boosting.plots package

Submodules

cyclic_boosting.plots.plot_utils module

cyclic_boosting.plots.plot_utils.add_missing_values_box(ax, nan_value=None, nan_uncert=None, nan_count=None)

Adds a Neurobayes analysis-plot-style box to plot axes indicating the corresponding value for NaN-data.

The function assumes that there is free space to the right of the plot. It was originally designed to fit the NaN-box between a 2-D matplotlib.pyplot.imshow() plot and its colorbar (see example below).

Parameters:

• ax (matplotlib.axes.Axes) – The axes object corresponding to the plot to which the nan box should be added

• nan_value (float) – A value to be displayed in the NaN box

• nan_uncert (float) – If not None, uncertainty that will be printed as plusminus deviation

Returns:

axes object of the NaN-box

Return type:

matplotlib.axes.Axes

cyclic_boosting.plots.plot_utils.append_extension(file, extension)

Small helper method to append possibly missing extension to a file name.

Only works on instances of basestring. Leaves all other objects unchanged

Parameters:

• file (any object) –

• extension (str) – extension (including ‘.’) to be appended to file

Returns:

result

Return type:

object of same type as file

cyclic_boosting.plots.plot_utils.blue_cyan_green_cmap()

Colormap from blue over cyan to green

Return type:

matplotlib.colors.LinearSegmentedColormap

cyclic_boosting.plots.plot_utils.calc_extent_with_missing_values_box(xmin, xmax)

Calculates new extent of a plot when a missing values box is added

Parameters:

• xmin (float) – minimum x-limit of the original plot

• xmax (float) – maximum x-limit of the original plot

Returns:

tuple of x-coordinates for the middle and the right edge of the box (the left edge is already given by xmax)

Return type:

(float, float)

cyclic_boosting.plots.plot_utils.colorful_histogram(ax, x, bin_boundaries, cmap)

Plot a histogram on a given matplotlib axis ax for to-be-binned values x with bin boundaries bin_boundaries (is passed to np.histogram). Each bin column is colored using the map provided by cmap.

cyclic_boosting.plots.plot_utils.fill_missing_values_box(ax, xmax, xmin, ymin, ymax)

Fill the yellow area and return x coordinates for the missing values box.

Parameters:

• ax (matplotlib.axes.Axes) – axes object into which the missing values box will be drawn

• xmax (float) – maximum x-limit of the axes

• xmin (float) – minimum x-limit of the axes

• ymin (float) – minimum y-limit of the axes

• ymax (float) – maximum y-limit of the axes

Note

Be careful of the unintuitive order of the parameters.

Returns:

tuple of x-coordinates for the middle and the right edge of the box (the left edge is already given by xmax)

Return type:

(float, float)

cyclic_boosting.plots.plot_utils.nbpy_style_context()

Activate nbpy’s Matplotlib style default settings locally.

Module contents

Plots for the Cyclic Boosting family

cyclic_boosting.plots.plot_analysis(plot_observer, file_obj, binners=None, figsize=(11.69, 8.27), use_tightlayout=True, plot_yp=True)

Plot factors as a multipage PDF, also include plots for Loss and factor change behaviour and an insample diagonal plot.

Parameters:

• plot_observer (PlottingObserver) – A fitted plotting observer.

• file (string or file-like) – If a string indicates the name of the file, to which the plots are written. The ending ‘.pdf’ is added if it is not present already. You may also pass a file-like object.

• binners (list) – A list of binners. If binners are given the labels of the x-axis of factor plots are better interpretable.

• figsize (tuple) – A tuple with length containing the width and height of the figures.

• use_tightlayout (bool) – If true the tightlayout option of matplotlib is used.

cyclic_boosting.plots.plot_factor_1d(feature, bin_bounds=None, with_errorbars=True, ylimits_include_errors=True, link_function=None, plot_yp=True)

Plots a single one dimensional factor plot.

Parameters:

• bin_bounds (list) – Bin boundaries to label the bins.

• feature (cyclic_boosting.base.Feature) – Feature as it can be obtained from the plotting observers features property.

• link_function (cyclic_boosting.link.LinkFunction) – Link function of the plotted feature

• with_errorbars (bool) – Option to switch errorbars on/off.

• ylimits_include_errors (bool) – Option to show the errorbars in the plot completely.

• plot_yp (bool) – Show deviation between truth and prediction in last iteration.

cyclic_boosting.plots.plot_factor_2d(n_bins_finite, feature, grid_item=None)

Plots a single two dimensional factor plot. For an example see the cyclic_boosting_analysis_plots

Parameters:

• n_bins_finite (int) – Number of finite bins

• feature (cyclic_boosting.base.Feature) – Feature as it can be obtained from the plotting observers features property.

• grid_item (matplotlib.gridspec.SubplotSpec) – If the plot should be imbedded into a larger grid.

cyclic_boosting.plots.plot_factor_change(plot_observer)

Plot the global factor changes for all iterations.

Parameters:

plot_observer (PlottingObserver) – A fitted plotting observer.

cyclic_boosting.plots.plot_factor_high_dim(feature)

Plots a histogram of the given factors with a logarithmic y-axis.

Parameters:

feature (cyclic_boosting.base.Feature) – Feature object from as it can be obtained from a plotting observer

cyclic_boosting.plots.plot_factor_histogram(feature)

Plots a histogram of the given factors with a logarithmic y-axis.

Parameters:

feature (cyclic_boosting.base.Feature) – Feature object from as it can be obtained from a plotting observer

cyclic_boosting.plots.plot_factors(plot_observer, binners=None, feature_groups_or_ids=None, features_per_row=2, use_tightlayout=True, plot_yp=True)

Create a matplotlib figure containing several factor plots (of a possibly pre-defined subset of features) in a grid.

Parameters:

• plot_observer (PlottingObserver) – A fitted plotting observer.

• binners (list) – A list of binners. If binners are given the labels of the x-axis of factor plots are better interpretable.

• feature_groups_or_ids (list) – A list of feature group names or cyclic_boosting.base.FeatureID for which the factors will be plotted. Default is to plot the factors of all features.

• features_per_row (int) – The number of factor plots in one row.

• use_tightlayout (bool) – If true the tightlayout option of matplotlib is used.

cyclic_boosting.plots.plot_in_sample_diagonal_plot(plot_observer)

Plot the in sample diagonal plot for prediction and truth.

Parameters:

plot_observer (PlottingObserver) – A fitted plotting observer.

cyclic_boosting.plots.plot_iteration_info(plot_observer)

Convenience method calling plot_loss() and plot_factor_change().

Parameters:

plot_observer (PlottingObserver) – A fitted plotting observer.

cyclic_boosting.plots.plot_loss(plot_observer)

Plot the change of the loss function between the in-sample y and the predicted factors for all iterations. For the zeroth iteration the mean of y is used as prediction.

Parameters:

plot_observer (PlottingObserver) – A fitted plotting observer.