python - 循环遍历 python 数组以匹配第二个数组中的多个条件，快速方法？

Question

我是 Python 的初学者，想知道是否有更快的方法来执行此代码，所以请原谅我的无知。我有 2 个 Excel 工作表:一个 (results) 有大约 30,000 行唯一用户 ID，然后我有 30 列问题，下面的单元格是空的。我的第二张工作表 (answers) 大约有 400,000 行和 3 列。第一列是用户 ID，第二列是提出的问题，第三列是用户对每个相应问题的回答。我想要做的本质上是一个索引匹配数组 excel 函数，我可以在其中通过匹配用户 ID 和询问的问题，用工作表 2 中的答案填充工作表 1 中的空白单元格。

现在我写了一段代码，但只处理工作表 1 中的 4 列就花了大约 2 个小时。我想弄清楚我的做法是否没有充分利用 Numpy 的功能。

import pandas as pd
import numpy as np

# Need to take in data from 'answers' and merge it into the 'results' data
# Will requiring matching the data based on 'id' in column 1 of 'answers' and the
# 'question' in column 2 of 'answers'
results = pd.read_excel("/Users/data.xlsx", 'Results')
answers = pd.read_excel("/Users/data.xlsx", 'Answers')

answers_array = np.array(answers) #########

# Create a list of questions being asked that will be matched to column 2 in answers. 
# Just getting all the questions I want
column_headers = list(results.columns)
formula_headers = []              #########
for header in column_headers:
   formula_headers.append(header)
del formula_headers[0:13]

# Create an empty array with ids in which the 'merged' data will be fed into
pre_ids = np.array(results['Id'])
ids = np.reshape(pre_ids, (pre_ids.shape[0], 1))
ids = ids.astype(str)

zero_array = np.zeros((ids.shape[0], len(formula_headers)))
ids_array = np.hstack((ids, zero_array))    ##########


for header in range(len(formula_headers)):
    question_index = formula_headers[header]
    for user in range(ids_array.shape[0]):
        user_index = ids_array[user, 0]
        location = answers_array[(answers_array[:, 0] == int(user_index)) & (answers_array[:, 1] == question_index)]
        # This location formula is what I feel is messing everything up,
        # or could be because of the nested loops
        # If can't find the user id and question in the answers array
        if location.size == 0:
            ids_array[user][header + 1] = ''
        else:
            row_location_1 = np.where(np.all(answers_array == location[0], axis=1))
            row_location = int(row_location_1[0][0])
            ids_array[user][header + 1] = answers_array[row_location][2]

print ids_array

Answer 1

我们可以只旋转第二个数据框，而不是用第二个数据框的信息填充第一个数据框。

answers.set_index(['id', 'question']).answer.unstack()

如果您需要行和列与results 数据框中的相同，您可以添加一个reindex_like 方法

answers.set_index(['id', 'question']).answer.unstack().reindex_like(results)

如果有重复

cols = ['id', 'question']
answers.drop_duplicates(cols).set_index(cols).answer.unstack()