def plot_forces(variables: pd.DataFrame) -> None:
forces = variables.drop(["Height", "Weight"], axis=1).melt()
forces[["type", "variable"]] = forces["variable"].str.split(expand=True)
tables.describe_table(forces,
groupby=["variable", "type"],
description="variables")
row_kwargs = dict(shorthand="variable", title=None, sort=forces_order)
column = alt.Column("type", title=None)
forces_plot = (plot_kde().facet(
data=forces.query("type != 'Imb'"),
row=alt.Row(header=alt.Header(labelAngle=0, labelAlign="left"),
**row_kwargs),
column=column,
).resolve_scale(y="independent").properties(bounds="flush"))
imb_plot = (plot_kde().facet(
data=forces.query("type == 'Imb'"),
row=alt.Row(header=alt.Header(labelFontSize=0), **row_kwargs),
column=column,
).resolve_scale(y="independent").properties(bounds="flush"))
plots = (forces_plot | imb_plot).configure_facet(spacing=5)
st.altair_chart(plots)
def show_continent_cases(df, label):
'''
Creates chart of new cases of each continents
'''
new_df = modify_data(df)
new_df = get_continent_values(df, label)
chart = alt.Chart(new_df).transform_filter(
alt.datum.location != 'World').mark_area().encode(
x=alt.X('date:T', title='Date'),
y=alt.Y(label + ':Q', title=None),
color='continent:N',
row=alt.Row('continent:N', title=label.replace('_', ' ').title()),
tooltip=[
alt.Tooltip('date:T'),
alt.Tooltip('continent', title='continent'),
alt.Tooltip(label + ':Q',
format=",.0f",
title=label.replace('_', ' '))
],
).configure_header(
titleColor='grey',
titleFontSize=13,
labelFontSize=12,
).configure_axis(labelFontSize=11, titleFontSize=13,
titleColor='grey').configure_axisX(
labelAngle=-30, ).configure_legend(
titleFontSize=13,
labelFontSize=12,
).properties(height=60).resolve_scale(
y='independent').interactive()
st.altair_chart(chart, use_container_width=True)
def ridge_plot(d, value, groupby, step=30, overlap=0.8, sort=None):
return (
alt.Chart(d)
.transform_joinaggregate(mean_value=f"mean({value})", groupby=[groupby])
.transform_bin(["bin_max", "bin_min"], value)
.transform_aggregate(
value="count()", groupby=[groupby, "mean_value", "bin_min", "bin_max"]
)
.transform_impute(
impute="value", groupby=[groupby, "mean_value"], key="bin_min", value=0
)
.mark_area(
interpolate="monotone", fillOpacity=0.8, stroke="lightgray", strokeWidth=0.5
)
.encode(
alt.X("bin_min:Q", bin="binned", title='activation', axis=alt.Axis(format='%', labelFlush=False)),
alt.Y("value:Q", scale=alt.Scale(range=[step, -step * overlap]), axis=None),
alt.Fill(
"mean_value:Q",
legend=None,
scale=alt.Scale(
domain=[d[value].max(), d[value].min()], scheme="redyellowblue"
),
),
alt.Row(
f"{groupby}:N",
title=None,
sort=alt.SortArray(sort) if sort else None,
header=alt.Header(labelAngle=0, labelAlign="right", format="%B"),
),
)
.properties(bounds="flush", height=step)
.configure_facet(spacing=0)
.configure_view(stroke=None)
)
def plot_shap_values(X: pd.DataFrame, model: dict) -> pd.DataFrame:
target = "EB mean force"
# st.pyplot(
# shap.summary_plot(shap.TreeExplainer(model[target], data=X).shap_values(X), X)
# )
shap_values = pd.DataFrame(shap.TreeExplainer(model[target],
data=X).shap_values(X),
columns=X.columns)
y_order = shap_values.abs().mean().nlargest(6).index.to_list()
shap_values = shap_values[y_order].melt()
# shap_values["rank"] = X.rank().melt()["value"].values
shap_values["Z-score"] = ((X[y_order] - X[y_order].mean()) /
X[y_order].std()).melt()["value"].clip(
-0.5, 0.5)
# dist = (
# alt.Chart(shap_values)
# .mark_circle(size=100)
# .encode(
# alt.X("value", title=None),
# alt.Y("variable", title=None, sort=y_order),
# alt.Color("Z-score", scale=alt.Scale(scheme="redblue", domain=[-2.5, 2.5])),
# )
# )
# rule = alt.Chart(pd.DataFrame([{'zero': 0}])).mark_rule().encode(alt.X('zero'))
stripplot = alt.Chart(shap_values, height=20, width=width).mark_circle(
size=100, clip=True).encode(
alt.Y(
'jitter:Q',
title=None,
axis=alt.Axis(values=[0],
ticks=False,
grid=False,
labels=False),
),
alt.X('value',
title="Shap value",
scale=alt.Scale(domain=[-.4, .4])),
alt.Color("Z-score",
scale=alt.Scale(scheme="redblue", domain=[-0.5, 0.5])),
alt.Row(
'variable',
title=None,
sort=y_order,
header=alt.Header(
labelAngle=0,
labelAlign='left',
),
),
).transform_calculate(
jitter='sqrt(-2*log(random()))*cos(2*PI*random())'
).configure_facet(spacing=0).configure_view(stroke=None)
st.altair_chart(stripplot)
def plot_error_dist(predictions: pd.DataFrame) -> None:
predictions_melted = predictions.melt(id_vars="target",
value_vars=["MAE", "MAPE"])
tables.describe_table(predictions_melted, groupby=["target", "variable"])
row_kwargs = dict(shorthand="target", title=None, sort=forces_order)
column = alt.Column("variable", title=None)
mae = (plot_kde().facet(
data=predictions_melted.query("variable == 'MAE'"),
row=alt.Row(header=alt.Header(labelAngle=0, labelAlign="left"),
**row_kwargs),
column=column,
).resolve_scale(y="independent").properties(bounds="flush"))
mape = (plot_kde().facet(
data=predictions_melted.query("variable == 'MAPE'"),
row=alt.Row(header=alt.Header(labelFontSize=0), **row_kwargs),
column=column,
).resolve_scale(y="independent").properties(bounds="flush"))
plots = (mae | mape).configure_facet(spacing=5)
st.altair_chart(plots)
def plot_targets(targets: pd.DataFrame) -> None:
targets_melted = targets.melt()
tables.describe_table(targets_melted, description="targets")
dist_plot = (plot_kde().facet(
data=targets_melted,
row=alt.Row(
"variable",
title=None,
header=alt.Header(labelAngle=0, labelAlign="left"),
),
).configure_facet(spacing=5).resolve_scale(y="independent").properties(
bounds="flush"))
st.altair_chart(dist_plot)
def grouped_single(df_name):
alt.themes.enable('dark')
df_match = pd.DataFrame(eval(df_name))
ht = re.sub(r"[\[\]\']", "",
str(df_name)).replace("grpd_",
"").replace("coi3_",
"").replace("coi4_", "")[:-3]
pdt = re.sub(r"[\[\]\']", "", str(df_name)).replace("grpd_", "").replace(
"coi3_", "").replace("coi4_", "").replace("Head", "")
pn = re.sub(r"[\[\]\'_]", "",
str(df_name)).replace("grpd", "").replace(ht,
"").replace(pdt, "")
pn = str(eval(pn))
df_match.columns = ['a', 'Explain', 'variable', 'value']
df_match.variable = [i.replace("percent_", "") for i in df_match.variable]
g = alt.Chart(df_match).mark_bar().encode(
x=alt.X('value:Q', title="% of Occurences"),
y=alt.Y('variable:O',
axis=alt.Axis(title=None,
titlePadding=20,
offset=15,
ticks=False,
minExtent=60,
labelLimit=100,
domain=False)),
color=alt.Color("variable:N",
scale=alt.Scale(scheme="inferno"),
legend=None),
row=alt.Row("a:N", title=ht_dict[ht] + " " + pdt_dict[pdt]),
tooltip=["Explain"]).configure(padding={
"left": 15,
"top": 25,
"right": 75,
"bottom": 25
}).configure_axisX(labelFontSize=15, titleFontSize=20).configure_axisY(
labelFontSize=15, titleFontSize=20).configure_title(
align="center", anchor="middle",
dx=50, fontSize=50).properties(title=re.sub(
r"[\[\]\']", "", str(pn)).replace(",", " "),
width=725)
g.save(
str("outputs/grouped_" + ht + pdt +
re.sub(r"[\[\]\' ]", "", str(pn)).replace(",", "") + ".html"))
return g
def generate_ridgeline_plot(data, attribute):
'''ridge line plot: multiple histograms overlaps'''
step = 40
overlap = 1
graph = alt.Chart(data).transform_joinaggregate(
mean_attribute="mean({})".format(str(attribute)),
groupby=['species']).transform_bin([
'bin_max', 'bin_min'
], str(attribute)).transform_aggregate(
value='count()',
groupby=[
'species', 'mean_attribute', 'bin_min', 'bin_max'
]).transform_impute(
impute='value',
groupby=['species', 'mean_attribute'],
key='bin_min',
value=0).mark_area(
interpolate='monotone',
fillOpacity=0.4,
stroke='lightgray',
strokeWidth=0.3).encode(
alt.X('bin_min:Q', bin='binned', title=str(attribute)),
alt.Y('value:Q',
scale=alt.Scale(range=[step, -step * overlap]),
axis=None),
alt.Fill('mean_attribute:Q',
legend=None,
scale=alt.Scale(domain=[30, 5],
scheme='redyellowblue')),
alt.Row(
'species:O',
title='Species',
header=alt.Header(
labelAngle=0,
labelAlign='right'))).properties(
bounds='flush',
title='Comparison: {}'.format(
str(metadata_description[attribute])),
height=100,
width=700,
).configure_facet(spacing=0, ).configure_view(
stroke=None, ).configure_title(
anchor='end')
return graph
def freq_missing() -> None:
log.info("collecting frequency counts")
o = Moby(ORIGINAL, ORIGINAL_TEXT_PATH, LIMITER)
search = set(["jonah", "bildad", "pip", "sperm", "right", "greenland"])
title = "_".join(sorted(search))
tag = "NNP"
data = alt.Data(values=(o.freq(search, tag)))
alt.Chart(data, width=1000).mark_bar().encode(
x=alt.X(
title="Chapter",
bin={"extent": [1, 135], "step": 1},
field="chapter",
type="quantitative",
),
y=alt.Y(title=f"word count", aggregate="count", type="quantitative"),
row=alt.Row(field="word", type="nominal"),
).save(f"{OUTPUT_DIR}/{title}_index_hist.html")
return None
