Data Frames Pandas 中所有行的 Pearson 相关性
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/37969282/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
Pearson correlation for all rows in Data Frames Pandas
提问by Batuhan B
I have a dataframe in Pandas and its shape is (136, 1445). I try to create correlation(Pearson) matrix for my 136 rows. So in the result, i need a matrix with size 136x136.
我在 Pandas 中有一个数据框，它的形状是 (136, 1445)。我尝试为我的 136 行创建相关（Pearson）矩阵。所以在结果中，我需要一个大小为 136x136 的矩阵。
I tried two different ways but i cannot get a results from them or when i create a 136x136 correlation matrix, i lost the columns name of dataframe.
我尝试了两种不同的方法，但无法从中获得结果，或者当我创建 136x136 相关矩阵时，我丢失了数据框的列名称。
First,
第一的，
gene_expression = pd.read_csv('padel_all_drug_results_original.csv',dtype='unicode')
gene_expression = gene_expression.convert_objects(convert_numeric=True)
gene_expression.corr()
This gives the column based pearson correlation matrix(1445*1445), and when I try to transpose my dataframe and then try to find correlation, the structure of dataframe is broken (like the columns name is lost or i dont even sure that the correlations are correct or not).
这给出了基于列的皮尔逊相关矩阵（1445 * 1445），当我尝试转置我的数据框然后尝试找到相关性时，数据框的结构被破坏（例如列名丢失或者我什至不确定相关性正确与否）。
Secondly,
其次，
distance = lambda column1, column2: pearsonr(column1,column2)[0]
result = gene_expression.apply(lambda col1: gene_expression.apply(lambda col2: distance(col1, col2)))
What should i do to calculate 136x136 pearson correlation matrix to do not change the original dataframe ?
我应该怎么做才能计算 136x136 皮尔逊相关矩阵以不更改原始数据帧？
Also, I have a 1445 features and some of columns nearly full of zeros. So I dropped that columns because they are noisy columns but have you got another idea to feature redeuction ?
此外，我有 1445 个特征和一些几乎全为零的列。所以我放弃了那些列，因为它们是嘈杂的列，但你有另一个想法来减少特征吗？
Thanks in advance
提前致谢
回答by Stefan
To get the correlation matrix containing pairwise correlation between all rows, you can:
要获得包含所有行之间成对相关的相关矩阵，您可以：
gene_expression.T.corr()
Using a toy example:
使用玩具示例：
df = pd.DataFrame(np.random.randint(0, high=100, size=(5, 10)), index=list(string.ascii_lowercase[:5]))
with 5 labeled rows and 10 columns:
带有 5 个标记的行和 10 列：
df.info()
Index: 5 entries, a to e
Data columns (total 10 columns):
0 5 non-null int64
1 5 non-null int64
2 5 non-null int64
3 5 non-null int64
4 5 non-null int64
5 5 non-null int64
6 5 non-null int64
7 5 non-null int64
8 5 non-null int64
9 5 non-null int64
dtypes: int64(10)
memory usage: 440.0+ bytes
Using
使用
df.T.corr()
yields
产量
a b c d e
a 1.000000 0.209460 -0.205302 -0.294427 0.353803
b 0.209460 1.000000 -0.530715 -0.117949 0.775848
c -0.205302 -0.530715 1.000000 -0.245101 -0.344358
d -0.294427 -0.117949 -0.245101 1.000000 0.058302
e 0.353803 0.775848 -0.344358 0.058302 1.000000
