def plotConfusionMatrix(df, width, height):
from bokeh.palettes import Blues8
# Had a specific mapper to map color with value
mapper = LinearColorMapper(palette=Blues8[::-1],
low=df.value.min(),
high=df.value.max())
TOOLS = "hover,save,reset"
# Define a figure
p = figure(
plot_width=width,
plot_height=height,
# title="",
x_range=list(df.Treatment.drop_duplicates()),
y_range=list(df.Prediction.drop_duplicates()),
toolbar_location='above',
tools=TOOLS,
tooltips=[('Counts', '@value')],
x_axis_location="below")
p.xaxis.axis_label = "Prediction"
p.yaxis.axis_label = "Truth"
# Create rectangle for heatmap
p.rect(x="Prediction",
y="Treatment",
width=1,
height=1,
source=ColumnDataSource(df),
line_color=None,
fill_color=transform('value', mapper))
# Add legend
color_bar = ColorBar(color_mapper=mapper,
location=(0, 0),
label_standoff=12,
border_line_color=None,
ticker=BasicTicker(desired_num_ticks=len(Blues8)))
color_bar.background_fill_alpha = 0.0
p.add_layout(color_bar, 'right')
p.background_fill_alpha = 0.0
p.border_fill_alpha = 0.0
return p
def bkplot(self,
x,
y,
color='None',
radii='None',
ps=20,
minps=0,
alpha=0.8,
pw=600,
ph=400,
palette='Inferno256',
style='smapstyle',
Hover=True,
title='',
table=False,
table_width=600,
table_height=150,
add_colorbar=True,
Periodic_color=False,
return_datasrc=False,
frac_load=1.0,
marker=['circle'],
seed=0,
**kwargs):
from bokeh.layouts import row, widgetbox, column, Spacer
from bokeh.models import HoverTool, TapTool, FixedTicker, Circle, WheelZoomTool
from bokeh.models import CustomJS, Slider, Rect, ColorBar, HoverTool, LinearColorMapper, BasicTicker
from bokeh.plotting import figure
import bokeh.models.markers as Bokeh_markers
from bokeh.models import ColumnDataSource, CDSView, IndexFilter
from bokeh.palettes import all_palettes, Spectral6, Inferno256, Viridis256, Greys256, Magma256, Plasma256
from bokeh.palettes import Spectral, Inferno, Viridis, Greys, Magma, Plasma
from bokeh.models import LogColorMapper, LogTicker, ColorBar, BasicTicker, LinearColorMapper
from bokeh.models.widgets import DataTable, TableColumn, NumberFormatter, Div
import pandas as pd
# if (title==''): title=self.name
fulldata = self.pd
idx = np.arange(len(fulldata))
fulldata['id'] = idx
nload = int(frac_load * len(fulldata))
np.random.seed(seed)
np.random.shuffle(idx)
idload = np.sort(idx[0:nload])
data = self.pd.iloc[idload].copy()
if palette == 'cosmo': COLORS = cosmo()
else: COLORS = locals()[palette]
marklist = [
'circle', 'diamond', 'triangle', 'square', 'asterisk', 'cross',
'inverted_triangle'
]
if not marker[0] in marklist: marker = marklist
# TOOLS="resize,crosshair,pan,wheel_zoom,reset,tap,save,box_select,box_zoom,lasso_select"
TOOLS = "pan,reset,tap,save,box_zoom,lasso_select"
wheel_zoom = WheelZoomTool(dimensions='both')
if Hover:
proplist = []
for prop in data.columns:
if prop not in [
"CV1", "CV2", "Cv1", "Cv2", "cv1", "cv2", "colors",
"radii", "id"
]:
proplist.append((prop, '@' + prop))
hover = HoverTool(names=["mycircle"], tooltips=[("id", '@id')])
for prop in proplist:
hover.tooltips.append(prop)
plot = figure(title=title,
plot_width=pw,
active_scroll=wheel_zoom,
plot_height=ph,
tools=[TOOLS, hover, wheel_zoom],
**kwargs)
else:
plot = figure(title=title,
plot_width=pw,
active_scroll=wheel_zoom,
plot_height=ph,
tools=[TOOLS],
**kwargs)
# selection glyphs and plot styles
mdict = {
'circle': 'Circle',
'diamond': 'Diamond',
'triangle': 'Triangle',
'square': 'Square',
'asterisk': 'Asterisk',
'cross': 'Cross',
'inverted_triangle': 'InvertedTriangle'
}
initial_circle = Circle(x='x', y='y')
selected_circle = getattr(Bokeh_markers,
mdict[marker[0]])(fill_alpha=0.7,
fill_color="blue",
size=ps * 1.5,
line_color="blue")
nonselected_circle = getattr(Bokeh_markers,
mdict[marker[0]])(fill_alpha=alpha * 0.5,
fill_color='colors',
line_color='colors',
line_alpha=alpha * 0.5)
# set up variable point size
if radii == 'None':
r = [ps for i in range(len(data))]
data['radii'] = r
else:
if data[radii].dtype == 'object': # Categorical variable for radii
grouped = data.groupby(radii)
i = 0
r = np.zeros(len(data))
for group_item in grouped.groups.keys():
r[grouped.groups[group_item].tolist()] = i**2
i = i + 2
else:
r = [val for val in data[radii]]
rn = self.normalize(r)
rad = [minps + ps * np.sqrt(val) for val in rn]
data['radii'] = rad
# setup variable point color
if color == 'None':
c = ["#31AADE" for i in range(len(data))]
data['colors'] = c
datasrc = ColumnDataSource(data)
getattr(plot, marker[0])(x,
y,
source=datasrc,
size='radii',
fill_color='colors',
fill_alpha=alpha,
line_color='colors',
line_alpha=alpha,
name="mycircle")
renderer = plot.select(name="mycircle")
renderer.selection_glyph = selected_circle
renderer.nonselection_glyph = nonselected_circle
else:
if data[color].dtype == 'object': # Categorical variable for colors
grouped = data.groupby(color)
# COLORS=Spectral[len(grouped)]
i = 0
nc = len(COLORS)
istep = int(nc / len(grouped))
cat_colors = []
for group_item in grouped.groups.keys():
# data.loc[grouped.groups[group_item],'colors']=COLORS[i]
# print(group_item,COLORS[i])
i = min(i + istep, nc - 1)
cat_colors.append(COLORS[i])
#colors=[ '#d53e4f', '#3288bd','#fee08b', '#99d594']
datasrc = ColumnDataSource(data)
view = []
# used_markers=[]
# marker=['circle','diamond','triangle','square','asterisk','cross','inverted_triangle']
#while True:
# for x in marker:
# used_markers.append(x)
# if len(used_markers)>len(grouped): break
i = 0
#print used_markers
for group_item in grouped.groups.keys():
view.append(
CDSView(
source=datasrc,
filters=[IndexFilter(grouped.groups[group_item])]))
cname = 'mycircle' + str(i)
#print used_markers[i]
try:
mk = marker[i]
except:
mk = marker[0]
getattr(plot, mk)(x,
y,
source=datasrc,
size='radii',
fill_color=cat_colors[i],
muted_color=cat_colors[i],
muted_alpha=0.2,
fill_alpha=alpha,
line_alpha=alpha,
line_color=cat_colors[i],
name=cname,
legend=group_item,
view=view[i])
selected_mk = getattr(Bokeh_markers,
mdict[mk])(fill_alpha=0.7,
fill_color="blue",
size=ps * 1.5,
line_color="blue",
line_alpha=0.7)
nonselected_mk = getattr(Bokeh_markers, mdict[mk])(
fill_alpha=alpha * 0.5,
fill_color=cat_colors[i],
line_color=cat_colors[i],
line_alpha=alpha * 0.5)
renderer = plot.select(name=cname)
renderer.selection_glyph = selected_mk
renderer.nonselection_glyph = nonselected_mk
i += 1
plot.legend.location = "top_left"
plot.legend.orientation = "vertical"
plot.legend.click_policy = "hide"
else:
if Periodic_color: # if periodic property then generate periodic color palatte
blendcolor = interpolate(COLORS[-1], COLORS[0],
len(COLORS) / 5)
COLORS = COLORS + blendcolor
groups = pd.cut(data[color].values, len(COLORS))
c = [COLORS[xx] for xx in groups.codes]
data['colors'] = c
datasrc = ColumnDataSource(data)
getattr(plot, marker[0])(x,
y,
source=datasrc,
size='radii',
fill_color='colors',
fill_alpha=alpha,
line_color='colors',
line_alpha=alpha,
name="mycircle")
renderer = plot.select(name="mycircle")
renderer.selection_glyph = selected_circle
renderer.nonselection_glyph = nonselected_circle
color_mapper = LinearColorMapper(COLORS,
low=data[color].min(),
high=data[color].max())
colorbar = ColorBar(color_mapper=color_mapper,
ticker=BasicTicker(),
label_standoff=4,
border_line_color=None,
location=(0, 0),
orientation="vertical")
colorbar.background_fill_alpha = 0
colorbar.border_line_alpha = 0
if add_colorbar:
plot.add_layout(colorbar, 'left')
# Overview plot
oplot = figure(title='',
plot_width=200,
plot_height=200,
toolbar_location=None)
oplot.circle(x,
y,
source=datasrc,
size=4,
fill_alpha=0.6,
line_color=None,
name="mycircle")
orenderer = oplot.select(name="mycircle")
orenderer.selection_glyph = selected_circle
# orenderer.nonselection_glyph = nonselected_circle
rectsource = ColumnDataSource({'xs': [], 'ys': [], 'wd': [], 'ht': []})
jscode = """
var data = source.data;
var start = range.start;
var end = range.end;
data['%s'] = [start + (end - start) / 2];
data['%s'] = [end - start];
source.change.emit();
"""
plot.x_range.callback = CustomJS(args=dict(source=rectsource,
range=plot.x_range),
code=jscode % ('xs', 'wd'))
plot.y_range.callback = CustomJS(args=dict(source=rectsource,
range=plot.y_range),
code=jscode % ('ys', 'ht'))
rect = Rect(x='xs',
y='ys',
width='wd',
height='ht',
fill_alpha=0.1,
line_color='black',
fill_color='red')
oplot.add_glyph(rectsource, rect)
# plot style
plot.toolbar.logo = None
oplot.toolbar.logo = None
if style == 'smapstyle': plist = [plot, oplot]
else: plist = [oplot]
for p in plist:
p.xgrid.grid_line_color = None
p.ygrid.grid_line_color = None
p.xaxis[0].ticker = FixedTicker(ticks=[])
p.yaxis[0].ticker = FixedTicker(ticks=[])
p.outline_line_width = 0
p.outline_line_alpha = 0
p.background_fill_alpha = 0
p.border_fill_alpha = 0
p.xaxis.axis_line_width = 0
p.xaxis.axis_line_color = "white"
p.yaxis.axis_line_width = 0
p.yaxis.axis_line_color = "white"
p.yaxis.axis_line_alpha = 0
# table
if table:
tcolumns = [
TableColumn(field='id',
title='id',
formatter=NumberFormatter(format='0'))
]
for prop in data.columns:
if prop not in [
"CV1", "CV2", "Cv1", "Cv2", "cv1", "cv2", "colors",
'id', "radii"
]:
if data[prop].dtype == 'object':
tcolumns.append(TableColumn(field=prop, title=prop))
if data[prop].dtype == 'float64':
tcolumns.append(
TableColumn(
field=prop,
title=prop,
formatter=NumberFormatter(format='0.00')))
if data[prop].dtype == 'int64':
tcolumns.append(
TableColumn(field=prop,
title=prop,
formatter=NumberFormatter(format='0')))
data_table = DataTable(source=datasrc,
fit_columns=True,
scroll_to_selection=True,
columns=tcolumns,
name="Property Table",
width=table_width,
height=table_height)
div = Div(text="""<h6><b> Property Table </b> </h6> <br>""",
width=600,
height=10)
if return_datasrc:
return plot, oplot, column(widgetbox(div), Spacer(height=10),
widgetbox(data_table)), datasrc
else:
return plot, oplot, column(widgetbox(div), Spacer(height=10),
widgetbox(data_table))
else:
return plot, oplot
