统计图表

直方图

使用 quad() 图形来创建从 np.histogram 输出绘制的直方图。

import numpy as np

from bokeh.plotting import figure, show

rng = np.random.default_rng()
x = rng.normal(loc=0, scale=1, size=1000)

p = figure(width=670, height=400, toolbar_location=None,
           title="Normal (Gaussian) Distribution")

# Histogram
bins = np.linspace(-3, 3, 40)
hist, edges = np.histogram(x, density=True, bins=bins)
p.quad(top=hist, bottom=0, left=edges[:-1], right=edges[1:],
         fill_color="skyblue", line_color="white",
         legend_label="1000 random samples")

# Probability density function
x = np.linspace(-3.0, 3.0, 100)
pdf = np.exp(-0.5*x**2) / np.sqrt(2.0*np.pi)
p.line(x, pdf, line_width=2, line_color="navy",
       legend_label="Probability Density Function")

p.y_range.start = 0
p.xaxis.axis_label = "x"
p.yaxis.axis_label = "PDF(x)"

show(p)

人口金字塔图是一种发散型水平条形图，可用于比较两个组之间的分布。在 Bokeh 中，可以使用 hbar() 图形来创建它们。

import numpy as np

from bokeh.models import CustomJSTickFormatter, Label
from bokeh.palettes import DarkText, Vibrant3 as colors
from bokeh.plotting import figure, show
from bokeh.sampledata.titanic import data as df

sex_group = df.groupby("sex")

female_ages = sex_group.get_group("female")["age"].dropna()
male_ages = sex_group.get_group("male")["age"].dropna()

bin_width = 5
bins = np.arange(0, 72, bin_width)
m_hist, edges = np.histogram(male_ages, bins=bins)
f_hist, edges = np.histogram(female_ages, bins=bins)

p = figure(title="Age population pyramid of titanic passengers, by gender", height=400, width=600,
           x_range=(-90, 90), x_axis_label="count")

p.hbar(right=f_hist, y=edges[1:], height=bin_width*0.8, color=colors[0], line_width=0)

p.hbar(right=m_hist * -1, y=edges[1:], height=bin_width*0.8, color=colors[1], line_width=0)

# add text to every other bar
for i, (count, age) in enumerate(zip(f_hist, edges[1:])):
    if i % 2 == 1:
        continue
    p.text(x=count, y=edges[1:][i], text=[f"{age-bin_width}-{age}yrs"],
           x_offset=5, y_offset=7, text_font_size="12px", text_color=DarkText[5])

# customise x-axis and y-axis
p.xaxis.ticker = (-80, -60, -40, -20, 0, 20, 40, 60, 80)
p.xaxis.major_tick_out = 0
p.y_range.start = 3
p.ygrid.grid_line_color = None
p.yaxis.visible = False

# format tick labels as absolute values for the two-sided plot
p.xaxis.formatter = CustomJSTickFormatter(code="return Math.abs(tick);")

# add labels
p.add_layout(Label(x=-40, y=70, text="Men", text_color=colors[1], x_offset=5))
p.add_layout(Label(x=20, y=70, text="Women", text_color=colors[0], x_offset=5))

show(p)

箱线图

箱线图可以使用 Whisker 标注、vbar() 和 scatter() 图形来组装。

import pandas as pd

from bokeh.models import ColumnDataSource, Whisker
from bokeh.plotting import figure, show
from bokeh.sampledata.autompg2 import autompg2
from bokeh.transform import factor_cmap

df = autompg2[["class", "hwy"]].rename(columns={"class": "kind"})

kinds = df.kind.unique()

# compute quantiles
qs = df.groupby("kind").hwy.quantile([0.25, 0.5, 0.75])
qs = qs.unstack().reset_index()
qs.columns = ["kind", "q1", "q2", "q3"]

# compute IQR outlier bounds
iqr = qs.q3 - qs.q1
qs["upper"] = qs.q3 + 1.5*iqr
qs["lower"] = qs.q1 - 1.5*iqr
df = pd.merge(df, qs, on="kind", how="left")

source = ColumnDataSource(qs)

p = figure(x_range=kinds, tools="", toolbar_location=None,
           title="Highway MPG distribution by vehicle class",
           background_fill_color="#eaefef", y_axis_label="MPG")

# outlier range
whisker = Whisker(base="kind", upper="upper", lower="lower", source=source)
whisker.upper_head.size = whisker.lower_head.size = 20
p.add_layout(whisker)

# quantile boxes
cmap = factor_cmap("kind", "TolRainbow7", kinds)
p.vbar("kind", 0.7, "q2", "q3", source=source, color=cmap, line_color="black")
p.vbar("kind", 0.7, "q1", "q2", source=source, color=cmap, line_color="black")

# outliers
outliers = df[~df.hwy.between(df.lower, df.upper)]
p.scatter("kind", "hwy", source=outliers, size=6, color="black", alpha=0.3)

p.xgrid.grid_line_color = None
p.axis.major_label_text_font_size="14px"
p.axis.axis_label_text_font_size="12px"

show(p)

核密度估计

import numpy as np
from scipy.stats import gaussian_kde

from bokeh.palettes import Blues9
from bokeh.plotting import figure, show
from bokeh.sampledata.autompg import autompg as df


def kde(x, y, N):
    xmin, xmax = x.min(), x.max()
    ymin, ymax = y.min(), y.max()

    X, Y = np.mgrid[xmin:xmax:N*1j, ymin:ymax:N*1j]
    positions = np.vstack([X.ravel(), Y.ravel()])
    values = np.vstack([x, y])
    kernel = gaussian_kde(values)
    Z = np.reshape(kernel(positions).T, X.shape)

    return X, Y, Z

x, y, z = kde(df.hp, df.mpg, 300)

p = figure(height=400, x_axis_label="hp", y_axis_label="mpg",
           background_fill_color="#fafafa", tools="", toolbar_location=None,
           title="Kernel density estimation plot of HP vs MPG")
p.grid.level = "overlay"
p.grid.grid_line_color = "black"
p.grid.grid_line_alpha = 0.05

palette = Blues9[::-1]
levels = np.linspace(np.min(z), np.max(z), 10)
p.contour(x, y, z, levels[1:], fill_color=palette, line_color=palette)

show(p)

核密度估计也可以使用 varea() 图形来绘制。

import numpy as np
from sklearn.neighbors import KernelDensity

from bokeh.models import ColumnDataSource, Label, PrintfTickFormatter
from bokeh.palettes import Dark2_5 as colors
from bokeh.plotting import figure, show
from bokeh.sampledata.cows import data as df

breed_groups = df.groupby('breed')
x = np.linspace(2, 8, 1000)
source = ColumnDataSource(dict(x=x))

p = figure(title="Multiple density estimates", height=300, x_range=(2.5, 7.5), x_axis_label="butterfat contents", y_axis_label="density")

for (breed, breed_df), color in zip(breed_groups, colors):
    data = breed_df['butterfat'].values
    kde = KernelDensity(kernel="gaussian", bandwidth=0.2).fit(data[:, np.newaxis])
    log_density = kde.score_samples(x[:, np.newaxis])
    y = np.exp(log_density)
    source.add(y, breed)
    p.varea(x="x", y1=breed, y2=0, source=source, fill_alpha=0.3, fill_color=color)

    # Find the highest point and annotate with a label
    max_idx = np.argmax(y)
    highest_point_label = Label(
        x=x[max_idx],
        y=y[max_idx],
        text=breed,
        text_font_size="10pt",
        x_offset=10,
        y_offset=-5,
        text_color=color,
    )
    p.add_layout(highest_point_label)

# Display x-axis labels as percentages
p.xaxis.formatter = PrintfTickFormatter(format="%d%%")

p.axis.axis_line_color = None
p.axis.major_tick_line_color = None
p.axis.minor_tick_line_color = None

p.xgrid.grid_line_color = None
p.yaxis.ticker = (0, 0.5, 1, 1.5)
p.y_range.start = 0

show(p)

Sina 图

Sina 图可以使用 harea() 和 scatter() 图形来组装。

import numpy as np
import pandas as pd
from sklearn.neighbors import KernelDensity

from bokeh.plotting import figure, show
from bokeh.sampledata.lincoln import data as df

df["DATE"] = pd.to_datetime(df["DATE"])
df["TAVG"] = (df["TMAX"] + df["TMIN"]) / 2
df["MONTH"] = df.DATE.dt.strftime("%b")

months = list(df.MONTH.unique())

p = figure(
    height=400,
    width=600,
    x_range=months,
    x_axis_label="month",
    y_axis_label="mean temperature (F)",
)

# add a non-uniform categorical offset to a given category
def offset(category, data, scale=7):
    return list(zip([category] * len(data), scale * data))


for month in months:
    month_df = df[df.MONTH == month].dropna()
    tavg = month_df.TAVG.values
    temps = np.linspace(tavg.min(), tavg.max(), 50)

    kde = KernelDensity(kernel="gaussian", bandwidth=3).fit(tavg[:, np.newaxis])
    density = np.exp(kde.score_samples(temps[:, np.newaxis]))
    x1, x2 = offset(month, density), offset(month, -density)

    p.harea(x1=x1, x2=x2, y=temps, alpha=0.8, color="#E0E0E0")

    # pre-compute jitter in Python, this case is too complex for BokehJS
    tavg_density = np.exp(kde.score_samples(tavg[:, np.newaxis]))
    jitter = (np.random.random(len(tavg)) * 2 - 1) * tavg_density

    p.scatter(x=offset(month, jitter), y=tavg, color="black")

p.y_range.start = -10
p.yaxis.ticker = [0, 25, 50, 75]
p.grid.grid_line_color = None

show(p)

SPLOM

SPLOM 是“散点图矩阵”，它以网格形式排列多个散点图，以突出显示各个维度之间的相关性。SPLOM 的关键组成部分是 链接平移 和 链接刷选，如本示例所示。

from itertools import product

from bokeh.io import show
from bokeh.layouts import gridplot
from bokeh.models import (BasicTicker, ColumnDataSource, DataRange1d,
                          Grid, LassoSelectTool, LinearAxis, PanTool,
                          Plot, ResetTool, Scatter, WheelZoomTool)
from bokeh.sampledata.penguins import data
from bokeh.transform import factor_cmap

df = data.copy()
df["body_mass_kg"] = df["body_mass_g"] / 1000

SPECIES = sorted(df.species.unique())
ATTRS = ("bill_length_mm", "bill_depth_mm", "body_mass_kg")
N = len(ATTRS)

source = ColumnDataSource(data=df)

xdrs = [DataRange1d(bounds=None) for _ in range(N)]
ydrs = [DataRange1d(bounds=None) for _ in range(N)]

plots = []

for i, (y, x) in enumerate(product(ATTRS, reversed(ATTRS))):
    p = Plot(x_range=xdrs[i%N], y_range=ydrs[i//N],
             background_fill_color="#fafafa",
             border_fill_color="white", width=200, height=200, min_border=5)

    if i % N == 0:  # first column
        p.min_border_left = p.min_border + 4
        p.width += 40
        yaxis = LinearAxis(axis_label=y)
        yaxis.major_label_orientation = "vertical"
        p.add_layout(yaxis, "left")
        yticker = yaxis.ticker
    else:
        yticker = BasicTicker()
    p.add_layout(Grid(dimension=1, ticker=yticker))

    if i >= N*(N-1):  # last row
        p.min_border_bottom = p.min_border + 40
        p.height += 40
        xaxis = LinearAxis(axis_label=x)
        p.add_layout(xaxis, "below")
        xticker = xaxis.ticker
    else:
        xticker = BasicTicker()
    p.add_layout(Grid(dimension=0, ticker=xticker))

    scatter = Scatter(x=x, y=y, fill_alpha=0.6, size=5, line_color=None,
                      fill_color=factor_cmap('species', 'Category10_3', SPECIES))
    r = p.add_glyph(source, scatter)
    p.x_range.renderers.append(r)
    p.y_range.renderers.append(r)

    # suppress the diagonal
    if (i%N) + (i//N) == N-1:
        r.visible = False
        p.grid.grid_line_color = None

    p.add_tools(PanTool(), WheelZoomTool(), ResetTool(), LassoSelectTool())

    plots.append(p)

show(gridplot(plots, ncols=N))