如何在直方图中绘制具有相同标签编号的多个特征

Leo 发表于 Dev

狮子座

我有一个包含8个功能和1个类的文本文件。我文件的数据是（data.txt）：

1,1,3,2,1,1,1,3,HIGH
1,1,3,1,2,1,1,3,HIGH
1,1,1,1,3,3,1,2,HIGH
1,3,2,1,3,3,3,3,HIGH
1,3,1,2,3,1,2,1,HIGH
2,3,1,2,1,2,2,1,HIGH
2,2,2,2,2,1,2,3,HIGH
2,2,1,1,1,2,2,3,HIGH
3,2,1,3,1,3,3,3,HIGH
3,2,1,2,2,3,3,2,HIGH

在上面的文件中，前8列是功能。它们用可能是1或2或3的数字标记。最后一列是类名（HIGH）。现在，我想根据标签号来绘制这些特征。我可以通过以下代码在第3列中进行操作：

import pandas as pd
from matplotlib import pyplot as plt

df = pd.read_csv('data.txt', header=None)

# Features are : A,B,C,...,H
df.columns = ['A', 'B','C', 'D', 'E', 'F', 'G', 'H', 'class']

X = df.ix[:, 0:8].values
y = df.ix[:, 8].values

kind = ['barstacked']
deg = ['HIGH']
pos = ['left','right','mid']
col = ['r','b','y']

with plt.style.context('seaborn-whitegrid'):
    plt.figure(figsize=(8, 6))

    for j in range(0,3):
        for i in range(1):
                plt.hist(X[y == deg[i], j],
                     label=deg[i],
                     bins=30,
                     alpha=0.6, histtype=kind[i], align=pos[j], color=col[j])

    plt.tick_params(axis='both', which='major', labelsize=17)
    plt.xlim(0.75, 3.25)
    plt.tight_layout()
    plt.savefig("figure.png" , format='png', dpi=700)
    plt.show()

结果如下：

However I could not plot the other 5 columns because I did not know how to put them next to each other as there are only 3 align options (left, mid and right). What I am looking at is a histogram plot for all 8 features that separates the features based on the tag number. A graph like this:

Sheldore

You don't need a histogram here and you can easily generate the required figure using a bar chart because you are just plotting a single frequency here. The idea is as follows:

Use Counter module from collections to get the frequency of 1, 2, and 3.
条形图的x位置将围绕1、2和3居中。但是，要获得理想的效果，可以通过偏移x位置来对其进行调整：将1、2、3和3的左边的前4个条形1、2、3右边的下4个小节。可以使用(j-4)*0.1添加到x值的offset参数来完成。在这里0.1是条形宽度的不错选择。
您无需在i此处进行额外的循环，因为它始终为0
df.ix在较新的熊猫版本中已弃用。您将不得不使用df.iloc。

以下是您的操作方法。

df.columns = ['A', 'B','C', 'D', 'E', 'F', 'G', 'H', 'class']

X = df.ix[:, 0:8].values
y = df.ix[:, 8].values

with plt.style.context('seaborn-whitegrid'):
    plt.figure(figsize=(8, 6))
    for j in range(0,8):
            freqs = Counter(X[y == deg[0], j])
            xvalues = np.array(list(freqs.keys()))
            plt.bar(xvalues+(j-4)*0.1, freqs.values(), width=0.1, 
                    alpha=0.9, edgecolor='k', lw=2)
    plt.tick_params(axis='both', which='major', labelsize=17)
    plt.xlim(0.25, 3.75)
    plt.xticks([1,2,3])
    plt.tight_layout()
    plt.show()