使用 pandas 过滤 csv 并创建矩阵的最快方法

Question

输入字典

{'basename_AM1.csv': ['AM1286', 'AM1287', 'AM1288']}

我有以下格式的大型 csv 文件 basename_AM1.csv 我有以下格式的大型 csv 文件

basename_AM1.csv

           ID1           ID2   Score
0        AM1287       AM1286  97.55
1        AM1288       AM1286  78.91
2        AM1289       AM1286  95.38
3        AM1290       AM1286  94.83
4        AM1291       AM1286  82.91

现在我需要通过搜索/过滤 csv 文件为给定的 input_dict 创建一个如下所示的相似性字典

{'AM1286': {'AM1286': 0, 'AM287': 97.55, 'AM288': 78.91},
 'AM1287': {'AM1286': 97.55, 'AM1287': 100.0, 'AM1288': 78.91},
 'AM1288': {'AM1286': 78.91, 'AM1287': 78.91, 'AM1288': 100.0}}

我提出了以下逻辑，但是对于 100 个样本的 input_dict，这需要很长时间，有人可以建议优化和最快的方法来实现这一点

for key,value in input_dict.items():
            base_name_df = pd.read_csv('csv_file_path')
            base_name_df.columns = "ID1","ID2","Score"
            if os.path.exists('csv_file_path'):
                for id1 in range(len(value)):
                    for id2 in range(len(value)):
                        scan_df = base_name_df[(base_name_df['ID1'] == value[id1]) & (base_name_df['ID2'] == value[id2])]
                        if not scan_df.empty:
                            scan_df = scan_df.groupby(['LIMSID1','LIMSID2'], as_index=False)['Score'].max()
                            final_dict[value[id1]][value[id2]] = scan_df.iloc[0]['Score']

Answer 1

IIUC，您可以使用：

input_dict = {'basename_AM1.csv': ['AM1286', 'AM1287', 'AM1288']}

import pandas as pd
for fname, lst in input_dict.items():
    df = pd.read_csv(fname, sep='\s+', names=['ID1', 'ID2', 'score'])
    df2 = df.pivot('ID1', 'ID2', 'score').reindex(index=lst, columns=lst)
    df2 = df2.combine_first(df2.T).fillna(0)
    # print for example
    print(df2.to_dict())

如果你想要 100 在对角线上：

import numpy as np
a = df2.to_numpy()
np.fill_diagonal(a, 100)
df2 = pd.DataFrame(a, index=lst, columns=lst)

输出：

{'AM1286': {'AM1286': 0.0, 'AM1287': 97.55, 'AM1288': 78.91},
 'AM1287': {'AM1286': 97.55, 'AM1287': 0.0, 'AM1288': 0.0},
 'AM1288': {'AM1286': 78.91, 'AM1287': 0.0, 'AM1288': 0.0}}

Answer 2

Pandas 有一个内置的 read_csv 方法。

文档位于： https ://pandas.pydata.org/docs/reference/api/pandas.read_csv.html