Average based on a Criteria/ Condition Numpy Python

Question

So I want to take the average of all values in column b when column a is a particular and plot it using Matplotlib.

So in the table above I want to average out the values in B and E for every same value in A and hence create a new element where

A =57 B= Avg of all values of b where A= 57 E= Avg of all values of e where A =57 and so on

And then finally plot the new element

I tried to implement it by taking the values into another Identity matrix but that does not work.

for x in list_of_entries:
    Final['A'] = x;
    Final['C'] = 0;
    Final['D'] = 1;
    I = np.logical_and((1), (data_temp['A'].astype(int) == x)) 
    Final['B'] =  np.average(data_temp[I]['B']);
    Final['E'] =  np.average(data_temp[I]['E']);
    np.empty(I);

Answer 1

仅使用NumPy，您可以使用np.unique(..., return_indx=True)查找用常量A值划分块的索引：

data_temp.sort(order=['A'])
uniqs, idx = np.unique(data_temp['A'], return_index=True)
idx = np.r_[idx, len(data_temp)]
# >>> idx
# array([  0,  10,  20,  33,  42,  50,  58,  71,  79,  90, 100])

然后，您可以使用

访问具有常量data_temp值的A块

data_temp[idx[i], idx[i+1]]

每个i = 0,..., len(idx)-1

。

这比使用

更快

for val in uniqs:
    mask = data_temp['A'] == val
    chunk = data_temp.loc[mask]

因为访问基本切片要比使用布尔选择掩码的高级索引快得多。

import numpy as np
import matplotlib.pyplot as plt
np.random.seed(2016)

data_temp = np.random.randint(10, size=(6*100)).view(
    [(col, '<i8')for col in list('ABCDEF')])
data_temp.sort(order=['A'])
uniqs, idx = np.unique(data_temp['A'], return_index=True)
idx = np.r_[idx, len(data_temp)]

result = []
for i in range(len(idx)-1):
    val = uniqs[i]
    start, end = idx[i], idx[i+1]
    # Uncomment to see the chunks of `data_temp` with constant A value
    # print(data_temp[start:end])
    mean = {col:data_temp[col][start:end].mean() for col in ['B', 'E']}
    result.append([val, mean['B'], 0, 1, mean['E']])
result = np.array(result)
print(result)

fig, ax = plt.subplots()
ax.plot(result[:, 0], result[:, 1])
ax.plot(result[:, 0], result[:, 4])
plt.show()

如果你有Pandas，整个计算变得非常简单：

import pandas as pd
import matplotlib.pyplot as plt
data_temp = pd.read_csv(dir_readfile, delimiter='\t', skiprows=1, names=names, 
    usecols=list(range(6)))
fig, ax = plt.subplots()
result = data_temp.groupby('A').agg({'B':'mean', 'E':'mean'})
result.plot()
plt.show()