如何在python中加速数据操作

Question

我有一个大文件，里面有2001-2017年的分钟级货币价格数据。我想创建一个简单的最近邻实现，看看5分钟前、25分钟前和50分钟前的%价格变化是否有任何解释能力(我相信它不会，但这只是为了学习)。我创建'dif_X‘列表的方式花费了非常长的时间(比如5个小时)。我对python非常陌生，甚至不知道在哪里可以找到解决这个问题的方法，但我知道有一种方法可以让它在python中运行得更快。代码如下：

import numpy as np
import pandas as pd

def findNNDistances(df_):
    samp = df_[10]

    count = 0
    df_['dist'] = [None]*len(df_)
    while count < len(df_):
        print("Count: " + str(count))
        df_['dist'] = np.sqrt((samp['dif_5'] - df_['dif_5'][count])**2 + 
                                  (samp['dif_25'] - df_['dif_25'][count])**2 + 
                                  (samp['dif_50'] - df_['dif_50'][count])**2)

df = pd.read_csv("Downloads/AUDUSD/AUDUSD.txt") # this is a csv

df['dif_5'] = [None]*len(df)
df['dif_25'] = [None]*len(df)
df['dif_50'] = [None]*len(df)
df['index'] = [None]*len(df)

count = 99
while count < len(df) - 1:
    print("countA: " + str(count))
    df['dif_5'][count] = (df['close'][count] - df['close'][count - 5])/df['close'][count-5]
    df['dif_25'][count] = (df['close'][count] - df['close'][count - 25])/df['close'][count-25]
    df['dif_50'][count] = (df['close'][count] - df['close'][count - 50])/df['close'][count-50] 
    df['index'][count] = count - 99
    count += 1

half_size = int(np.round(len(df)/2))
train = df[99:half_size] # not used yet
test = df[half_size + 1: len(df) - 1] # not used yet
df.apply(findNNDistances)
print(df['dist'].head(20))

怎样才能让它运行得更快呢？我也很欣赏一些关于在python中让这样的东西运行得更快的一般技巧。谢谢。