将日期时间与Seaborn的重绘一起使用 [英] Using Datetimes with Seaborn's Regplot

问题描述

我在Jupyter/IPython中工作，以绘制每天的单词数量，但是在Seaborn中将日期时间与Regplot一起使用时遇到了麻烦.尽管我正在尝试完成自动绘图，但不支持对日期数据进行回归不一定需要Regplot的解决方法-也许只是格式化x轴标签的一种方式.

I'm working in Jupyter/IPython to plot an amount of Words per Day, but am having trouble using datetimes with Regplot in Seaborn. Regplot by itself apparently does not support regression against date data, though what I am trying to accomplish does not necessarily require a workaround for Regplot - perhaps just a way of formatting the x-axis labels.

使用简单时间戳的最小工作示例:

A minimal working example, using simple timestamps:

%matplotlib inline
import numpy as np
import pandas as pd
import matplotlib as mpl
import matplotlib.pyplot as plt 
import matplotlib.dates as dates
import seaborn as sns
import time
import datetime
import radar
sns.set(style="whitegrid", color_codes=True)

data = pd.DataFrame([])

for i in np.arange(1, 10):
    date =  radar.random_datetime(start='2016-05-20', stop='2016-05-25')
    data = data.append(pd.DataFrame({'Date': time.mktime(date.timetuple()), 'Words': i + 100}, index=[0]), ignore_index=True)

points = plt.scatter(x = data['Date'], y = data["Words"], c=data["Words"], s=75, cmap="BrBG")
plt.colorbar(points)
sns.regplot(x = data['Date'], y = data["Words"], data=data, scatter=False, color='r')

哪一个用叠加的趋势线绘制散点图:

Which renders a scatterplot with an overlaid trend-line:

但是将日期作为日期时间:

But with the dates as datetimes:

points = plt.scatter(x = pd.to_datetime(data['Date'], unit='s').dt.to_pydatetime(), y = data["Words"], c=data["Words"], s=75, cmap="BrBG")
plt.colorbar(points)
sns.regplot(x = pd.to_datetime(data['Date'], unit='s').dt.to_pydatetime(), y = data["Words"], data=data, scatter=False, color='r')

它返回以下错误:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-7-d6488afe3dcb> in <module>()
      1 points = plt.scatter(x = pd.to_datetime(data['Date'], unit='s').dt.to_pydatetime(), y = data["Words"], c=data["Words"], s=75, cmap="BrBG")
      2 plt.colorbar(points)
----> 3 sns.regplot(x = pd.to_datetime(data['Date'], unit='s').dt.to_pydatetime(), y = data["Words"], data=data, scatter=False, color='r')

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in regplot(x, y, data, x_estimator, x_bins, x_ci, scatter, fit_reg, ci, n_boot, units, order, logistic, lowess, robust, logx, x_partial, y_partial, truncate, dropna, x_jitter, y_jitter, label, color, marker, scatter_kws, line_kws, ax)
    777     scatter_kws["marker"] = marker
    778     line_kws = {} if line_kws is None else copy.copy(line_kws)
--> 779     plotter.plot(ax, scatter_kws, line_kws)
    780     return ax
    781 

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in plot(self, ax, scatter_kws, line_kws)
    330             self.scatterplot(ax, scatter_kws)
    331         if self.fit_reg:
--> 332             self.lineplot(ax, line_kws)
    333 
    334         # Label the axes

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in lineplot(self, ax, kws)
    375 
    376         # Fit the regression model
--> 377         grid, yhat, err_bands = self.fit_regression(ax)
    378 
    379         # Get set default aesthetics

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in fit_regression(self, ax, x_range, grid)
    207             yhat, yhat_boots = self.fit_logx(grid)
    208         else:
--> 209             yhat, yhat_boots = self.fit_fast(grid)
    210 
    211         # Compute the confidence interval at each grid point

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in fit_fast(self, grid)
    222         grid = np.c_[np.ones(len(grid)), grid]
    223         reg_func = lambda _x, _y: np.linalg.pinv(_x).dot(_y)
--> 224         yhat = grid.dot(reg_func(X, y))
    225         if self.ci is None:
    226             return yhat, None

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\seaborn\linearmodels.py in <lambda>(_x, _y)
    221         X, y = np.c_[np.ones(len(self.x)), self.x], self.y
    222         grid = np.c_[np.ones(len(grid)), grid]
--> 223         reg_func = lambda _x, _y: np.linalg.pinv(_x).dot(_y)
    224         yhat = grid.dot(reg_func(X, y))
    225         if self.ci is None:

C:\Python\WinPython-64bit-3.5.2.2Qt5\python-3.5.2.amd64\lib\site-packages\numpy\linalg\linalg.py in pinv(a, rcond)
   1614     a, wrap = _makearray(a)
   1615     _assertNoEmpty2d(a)
-> 1616     a = a.conjugate()
   1617     u, s, vt = svd(a, 0)
   1618     m = u.shape[0]

AttributeError: 'datetime.datetime' object has no attribute 'conjugate'

尽管散点图确实以格式正确的日期时间呈现:

Though the scatterplot does render with the datetimes well-formatted:

是否可以通过Regplot使用日期时间或使用时间戳，但将x轴上的标签格式化为日期?

Is there a way to use datetimes with Regplot, or to use the timestamps but format the labels on the x-axis as dates?

推荐答案

您可以在xticks的位置获取时间戳的值，然后将其转换为所需的格式.

You can get the values of the timestamps at the locations of the xticks, and then convert them to your desired format.

ax = plt.gca()
xticks = ax.get_xticks()
xticks_dates = [datetime.datetime.fromtimestamp(x).strftime('%Y-%m-%d %H:%M:%S') for x in xticks]
ax.set_xticklabels(xticks_dates)

这篇关于将日期时间与Seaborn的重绘一起使用的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！