Visualization methods
Methods for visualizing evaluations results:
Radar Plot
evaluate.visualization.radar_plot
< source >( data model_names invert_range = [] config = None fig = None )
Parameters
- data (
List[dict]
) — the results (list of metric + value pairs). E.g. data = [{“accuracy”: 0.9, “precision”:0.8},{“accuracy”: 0.7, “precision”:0.6}] - names (
List[dict]
) — model names. E.g. names = [“model1”, “model 2”, …] - invert_range (
List[dict]
, optional) — the metrics to invert (in cases when smaller is better, e.g. speed) E.g. invert_range=[“latency_in_seconds”] - config (
dict
, optional) — a specification of the formatting configurations, namely:-
rad_ln_args (
dict
, default{"visible": True}
): The visibility of the radial (circle) lines. -
outer_ring (
dict
, default{"visible": True}
): The visibility of the outer ring. -
angle_ln_args (
dict
, default{"visible": True}
): The visibility of the angle lines. -
rgrid_tick_lbls_args (
dict
, default{"fontsize": 12}
): The font size of the tick labels on the scales. -
theta_tick_lbls (
dict
, default{"fontsize": 12}
): The font size of the variable labels on the plot. -
theta_tick_lbls_pad (
int
, default3
): The padding of the variable labels on the plot. -
theta_tick_lbls_brk_lng_wrds (
bool
, defaultTrue
): Whether long words in the label are broken up or not. -
theta_tick_lbls_txt_wrap (
int
, default15
): Text wrap for tick labels -
incl_endpoint (
bool
, defaultFalse
): Include value endpoints on calse -
marker (
str
, default"o"
): the shape of the marker used in the radar plot. -
markersize (
int
, default3
): the shape of the marker used in the radar plot. -
legend_loc (
str
, default"upper right"
): the location of the legend in the radar plot. Must be one of: ‘upper left’, ‘upper right’, ‘lower left’, ‘lower right’. -
bbox_to_anchor (
tuple
, default(2, 1)
: anchor for the legend.
-
- fig (
matplotlib.figure.Figure
, optional) — figure used to plot the radar plot.
Create a complex radar chart with different scales for each variable Source: https://towardsdatascience.com/how-to-create-and-visualize-complex-radar-charts-f7764d0f3652