gpt4 book ai didi

python - 根据整行屏蔽 Pandas DataFrame 行

转载 作者:太空宇宙 更新时间:2023-11-04 03:46:06 24 4
gpt4 key购买 nike

背景:

我正在处理 8 波段多光谱卫星图像,并根据反射率值估算水深。使用 statsmodels,我提出了一个 OLS 模型,该模型将根据每个像素的 8 个反射率值预测每个像素的深度。为了轻松地使用 OLS 模型,我将所有像素反射率值粘贴到格式如下例所示的 pandas 数据框中;其中每一行代表一个像素,每一列是多光谱图像的光谱带。

由于一些预处理步骤,所有岸上像素都已转换为全零。我不想尝试预测这些像素的“深度”,所以我想将我的 OLS 模型预测限制为不全为零值的行。

我需要将我的结果 reshape 为原始图像的行 x 列尺寸,这样我就不能只删除所有零行。

具体问题:

我有一个 Pandas 数据框。有些行包含全零。我想为某些计算屏蔽这些行,但我需要保留这些行。我不知道如何屏蔽全为零的行的所有条目。

例如:

In [1]: import pandas as pd
In [2]: import numpy as np
# my actual data has about 16 million rows so
# I'll simulate some data for the example.
In [3]: cols = ['band1','band2','band3','band4','band5','band6','band7','band8']
In [4]: rdf = pd.DataFrame(np.random.randint(0,10,80).reshape(10,8),columns=cols)
In [5]: zdf = pd.DataFrame(np.zeros( (3,8) ),columns=cols)
In [6]: df = pd.concat((rdf,zdf)).reset_index(drop=True)
In [7]: df
Out[7]:
band1 band2 band3 band4 band5 band6 band7 band8
0 9 9 8 7 2 7 5 6
1 7 7 5 6 3 0 9 8
2 5 4 3 6 0 3 8 8
3 6 4 5 0 5 7 4 5
4 8 3 2 4 1 3 2 5
5 9 7 6 3 8 7 8 4
6 6 2 8 2 2 6 9 8
7 9 4 0 2 7 6 4 8
8 1 3 5 3 3 3 0 1
9 4 2 9 7 3 5 5 0
10 0 0 0 0 0 0 0 0
11 0 0 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0

[13 rows x 8 columns]

我知道这样做可以得到我感兴趣的行:

In [8]: df[df.any(axis=1)==True]
Out[8]:
band1 band2 band3 band4 band5 band6 band7 band8
0 9 9 8 7 2 7 5 6
1 7 7 5 6 3 0 9 8
2 5 4 3 6 0 3 8 8
3 6 4 5 0 5 7 4 5
4 8 3 2 4 1 3 2 5
5 9 7 6 3 8 7 8 4
6 6 2 8 2 2 6 9 8
7 9 4 0 2 7 6 4 8
8 1 3 5 3 3 3 0 1
9 4 2 9 7 3 5 5 0

[10 rows x 8 columns]

但我需要稍后再次 reshape 数据,所以我需要将这些行放在正确的位置。我已经尝试了各种方法,包括 df.where(df.any(axis=1)==True) 但我找不到任何有效的方法。

失败:

  1. df.any(axis=1)==True 为我感兴趣的行提供 TrueFalse 对于我想屏蔽的行,但是当我尝试 df.where(df.any(axis=1)==True) 时,我只是取回了包含所有零的整个数据框。我想要整个数据框,但那些零行中的所有值都被屏蔽了,所以据我所知,它们应该显示为 Nan,对吧?

  2. 我尝试获取全为零的行的索引并按行屏蔽:

    mskidxs = df[df.any(axis=1)==False].index
    df.mask(df.index.isin(mskidxs))

    那也不管用,它给了我:

    ValueError: Array conditional must be same shape as self

    .index 只是返回一个 Int64Index。我需要一个与我的数据框尺寸相同的 bool 数组,但我不知道如何获得一个。

预先感谢您的帮助。

-贾里德

最佳答案

澄清我的问题的过程以迂回的方式引导我找到答案。 This question也帮助我指明了正确的方向。这是我想出来的:

import pandas as pd
# Set up my fake test data again. My actual data is described
# in the question.
cols = ['band1','band2','band3','band4','band5','band6','band7','band8']
rdf = pd.DataFrame(np.random.randint(0,10,80).reshape(10,8),columns=cols)
zdf = pd.DataFrame(np.zeros( (3,8) ),columns=cols)
df = pd.concat((zdf,rdf)).reset_index(drop=True)

# View the dataframe. (sorry about the alignment, I don't
# want to spend the time putting in all the spaces)
df

band1 band2 band3 band4 band5 band6 band7 band8
0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0
3 6 3 7 0 1 7 1 8
4 9 2 6 8 7 1 4 3
5 4 2 1 1 3 2 1 9
6 5 3 8 7 3 7 5 2
7 8 2 6 0 7 2 0 7
8 1 3 5 0 7 3 3 5
9 1 8 6 0 1 5 7 7
10 4 2 6 2 2 2 4 9
11 8 7 8 0 9 3 3 0
12 6 1 6 8 2 0 2 5

13 rows × 8 columns

# This is essentially the same as item #2 under Fails
# in my question. It gives me the indexes of the rows
# I want unmasked as True and those I want masked as
# False. However, the result is not the right shape to
# use as a mask.
df.apply( lambda row: any([i<>0 for i in row]),axis=1 )
0 False
1 False
2 False
3 True
4 True
5 True
6 True
7 True
8 True
9 True
10 True
11 True
12 True
dtype: bool

# This is what actually works. By setting broadcast to
# True, I get a result that's the right shape to use.
land_rows = df.apply( lambda row: any([i<>0 for i in row]),axis=1,
broadcast=True )

land_rows

Out[92]:
band1 band2 band3 band4 band5 band6 band7 band8
0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0
3 1 1 1 1 1 1 1 1
4 1 1 1 1 1 1 1 1
5 1 1 1 1 1 1 1 1
6 1 1 1 1 1 1 1 1
7 1 1 1 1 1 1 1 1
8 1 1 1 1 1 1 1 1
9 1 1 1 1 1 1 1 1
10 1 1 1 1 1 1 1 1
11 1 1 1 1 1 1 1 1
12 1 1 1 1 1 1 1 1

13 rows × 8 columns

# This produces the result I was looking for:
df.where(land_rows)

Out[93]:
band1 band2 band3 band4 band5 band6 band7 band8
0 NaN NaN NaN NaN NaN NaN NaN NaN
1 NaN NaN NaN NaN NaN NaN NaN NaN
2 NaN NaN NaN NaN NaN NaN NaN NaN
3 6 3 7 0 1 7 1 8
4 9 2 6 8 7 1 4 3
5 4 2 1 1 3 2 1 9
6 5 3 8 7 3 7 5 2
7 8 2 6 0 7 2 0 7
8 1 3 5 0 7 3 3 5
9 1 8 6 0 1 5 7 7
10 4 2 6 2 2 2 4 9
11 8 7 8 0 9 3 3 0
12 6 1 6 8 2 0 2 5

13 rows × 8 columns

再次感谢那些提供帮助的人。希望我找到的解决方案在某个时候对某些人有用。

我找到了另一种方法来做同样的事情。涉及的步骤更多,但根据 %timeit,它快了大约 9 倍。在这里:

def mask_all_zero_rows_numpy(df):
"""
Take a dataframe, find all the rows that contain only zeros
and mask them. Return a dataframe of the same shape with all
Nan rows in place of the all zero rows.
"""
no_data = -99
arr = df.as_matrix().astype(int16)
# make a row full of the 'no data' value
replacement_row = np.array([no_data for x in range(arr.shape[1])], dtype=int16)
# find out what rows are all zeros
mask_rows = ~arr.any(axis=1)
# replace those all zero rows with all 'no_data' rows
arr[mask_rows] = replacement_row
# create a masked array with the no_data value masked
marr = np.ma.masked_where(arr==no_data,arr)
# turn masked array into a data frame
mdf = pd.DataFrame(marr,columns=df.columns)
return mdf

mask_all_zero_rows_numpy(df)的结果应该和上面的Out[93]:一样。

关于python - 根据整行屏蔽 Pandas DataFrame 行,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/23798961/

24 4 0
Copyright 2021 - 2024 cfsdn All Rights Reserved 蜀ICP备2022000587号
广告合作:1813099741@qq.com 6ren.com