def populate_churn_plots(q):
shap_plot = churn_predictor.get_shap_explanation(
q.client.selected_customer_index)
q.page["shap_plot"] = ui.image_card(
box=config.boxes["shap_plot"],
title="",
type="png",
image=get_image_from_matplotlib(shap_plot),
)
top_negative_pd_plot = churn_predictor.get_top_negative_pd_explanation(
q.client.selected_customer_index)
q.page["top_negative_pd_plot"] = ui.image_card(
box=config.boxes["top_negative_pd_plot"],
title="Feature Most Contributing to Retention",
type="png",
image=get_image_from_matplotlib(top_negative_pd_plot),
)
top_positive_pd_plot = churn_predictor.get_top_positive_pd_explanation(
q.client.selected_customer_index)
q.page["top_positive_pd_plot"] = ui.image_card(
box=config.boxes["top_positive_pd_plot"],
title="Feature Most Contributing to Churn",
type="png",
image=get_image_from_matplotlib(top_positive_pd_plot),
)
def render_text_word_cloud_image(q: Q, image):
q.page['all'] = ui.image_card(
box=config.boxes['middle_panel'],
title=f'Word Cloud of the {config.column_mapping[q.client.review]}',
type='png',
image=image,
)
def render_all_text_word_cloud(q: Q):
image = plot_word_cloud(merge_to_single_text(config.dataset[q.client.review]))
q.page['all'] = ui.image_card(
box=config.boxes['middle_panel'],
title=f'Word Cloud of the {config.column_mapping[q.client.review]}',
type='png',
image=image,
)
def init(q: Q):
q.page['meta'] = ui.meta_card(
box='',
title='Explain Ratings',
layouts=[
ui.layout(breakpoint='xs',
zones=[
ui.zone('header'),
ui.zone('body',
direction=ui.ZoneDirection.ROW,
size='calc(100vh - 70px)',
zones=[
ui.zone('sidebar', size='350px'),
ui.zone('content',
direction=ui.ZoneDirection.ROW)
])
]),
])
q.page['header'] = ui.header_card(box='header',
title='Hotel Reviews',
subtitle='Explains the hotel reviews',
icon='ReviewSolid',
icon_color='#00A8E0')
q.client.review = config.review_column_list[0]
form_filters = []
for column in config.filterable_columns:
choices = [ui.choice(name='empty', label='All')] + [
ui.choice(name=str(column), label=str(column))
for column in config.dataset[column].drop_duplicates()
]
form_filters.append(
ui.dropdown(name=f'filter_{column}',
label=config.column_mapping[column],
trigger=True,
value='empty',
choices=choices))
sidebar_items = [
ui.dropdown(name='review',
label='Review type',
value=q.client.review,
trigger=True,
choices=[
ui.choice(name=column,
label=config.column_mapping[column])
for column in config.dataset[config.review_column_list]
]),
ui.separator('Filters')
] + form_filters
q.page['sidebar'] = ui.form_card(box='sidebar', items=sidebar_items)
q.page['original'] = ui.image_card(box='content',
title='Original',
type='png',
image=plot_word_cloud(
config.dataset[q.client.review], q))
def image_card_from_image(path_image: Path,
box: str,
title: str,
image_type: str = 'png') -> ui.image_card:
"""
Putting an image into Q ui.image_card.
"""
with open(Path(path_image), 'rb') as file_image:
base64_encoded_string = base64.b64encode(
file_image.read()).decode('utf8')
card = ui.image_card(
box=box,
title=title,
type=image_type,
image=base64_encoded_string,
)
return card
def render_compare_word_cloud(q: Q):
df = filter_data_frame(config.dataset, q.client.filters)
if len(df):
image = plot_word_cloud(merge_to_single_text(df[q.client.review]))
q.page['compare'] = ui.image_card(
box=config.boxes['right_panel'],
title='Word Cloud based on the selected filters',
type='png',
image=image,
)
else:
q.page['compare'] = ui.form_card(
box=config.boxes['right_panel'],
items=[
ui.message_bar(type='warning',
text='No reviews matching filter criteria!')
])
def render_customer_page(q: Q):
init(q)
selected_row = int(q.args.risk_table[0])
training_df = predictor.get_testing_data_as_pd_frame()
predictions_df = predictor.predicted_df.as_data_frame()
contributions_df = predictor.contributions_df.as_data_frame()
del contributions_df['BiasTerm']
q.client.selected_customer_id = training_df.loc[selected_row]["ID"]
print("selected id : ", q.client.selected_customer_id)
score = predictions_df.loc[selected_row]["predict"]
approve = bool(score < config.approval_threshold)
drop_column_from_df(training_df, 'default.payment.next.month')
add_column_to_df(training_df, predictions_df, 'Default Prediction Rate',
'predict')
training_df = round_df_column(training_df, 'Default Prediction Rate', 4)
render_customer_details_table(q, training_df, selected_row)
shap_plot = predictor.get_shap_explanation(selected_row)
q.page["shap_plot"] = ui.image_card(
box='shap_plot',
title="Effectiveness of each attribute on defaulting next payment",
type="png",
image=get_image_from_matplotlib(shap_plot, dpi=85),
)
render_customer_summary(q, training_df, contributions_df, selected_row,
approve)
q.page["buttons"] = ui.form_card(box='button_group',
items=[
ui.buttons([
ui.button(name='reject_btn',
label='Reject',
primary=not approve),
ui.button(name='approve_btn',
label='Approve',
primary=approve),
])
])
def render_diff_word_cloud(q: Q):
df = config.dataset
for key, value in q.client.filters.items():
df = df[df[key] == value]
if len(df):
q.page['diff'] = ui.image_card(box='content',
title='Diff',
type='png',
image=plot_word_cloud(
df[q.client.review], q))
else:
# TODO: Move into sidebar when https://github.com/h2oai/wave/pull/507 merged.
q.page['diff'] = ui.form_card(
box='content',
items=[
ui.message_bar(type='warning',
text='No reviews matching filter criteria!')
])
from h2o_wave import site, ui
import io
import base64
import numpy as np
import matplotlib.pyplot as plt
np.random.seed(19680801)
n = 25
plt.figure(figsize=(3, 3))
plt.scatter(
np.random.rand(n), np.random.rand(n),
s=(30 * np.random.rand(n)) ** 2,
c=np.random.rand(n),
alpha=0.5,
)
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
image = base64.b64encode(buf.read()).decode('utf-8')
page = site['/demo']
page['example'] = ui.image_card(
box='1 1 3 5',
title='An image',
type='png',
image=image,
)
page.save()
plt.scatter(
np.random.rand(n), np.random.rand(n),
s=(30 * np.random.rand(n)) ** 2,
c=np.random.rand(n),
alpha=0.5,
)
buf = io.BytesIO()
plt.savefig(buf, format='png')
buf.seek(0)
image = base64.b64encode(buf.read()).decode('utf-8')
page = site['/demo']
page['example1'] = ui.image_card(
box='1 1 3 5',
title='An image',
type='png',
image=image,
)
# Another way to achieve the same result is to use a data URL for the path:
# The example below constructs the data URL from the base64-encoded
# used in the previous example.
page['example2'] = ui.image_card(
box='1 6 3 5',
title='An image',
path=f"data:image/png;base64,{image}",
)
page.save()