我在 Python 中有以下代碼：import pandas as pdimport numpy as npdate_rng = pd.date_range(start='5/18/2019', end='7/22/2020', freq='S')df = pd.DataFrame(date_rng, columns=['start_timestamp'])df['end_timestamp'] = date_rngdf['start_timestamp'] = np.random.randint(1589760000,1595376000,size=(len(date_rng)))df['end_timestamp'] = np.random.randint(1589760000,1595376000,size=(len(date_rng)))df = df[(df.end_timestamp/df.start_timestamp<=1.000009)&(df.end_timestamp/df.start_timestamp>=1.000001)]df = df.sort_values(by=['start_timestamp','end_timestamp'])df['start_timestamp'] = pd.to_datetime(df['start_timestamp'],unit='s')df['end_timestamp'] = pd.to_datetime(df['end_timestamp'],unit='s')結(jié)果，我有以下數(shù)據(jù)框： start_timestamp end_timestamp2020-05-18 00:00:30 2020-05-18 00:54:072020-05-18 00:01:40 2020-05-18 03:50:392020-05-18 00:02:08 2020-05-18 02:39:412020-05-18 00:04:01 2020-05-18 00:47:252020-05-18 00:04:01 2020-05-18 02:26:502020-05-18 00:04:44 2020-05-18 02:17:53 . . .我應(yīng)該怎么做才能確保在我的數(shù)據(jù)集中每個end_timestamp都是在其下一行之前的日期時間start_timestamp？已實(shí)施的解決方案我基本上將數(shù)據(jù)集轉(zhuǎn)換為數(shù)組，將其按升序排列并將其轉(zhuǎn)換回?cái)?shù)據(jù)框。它可能不是最優(yōu)雅的解決方案，但它工作正常并為我打算使用的內(nèi)容生成了一致的數(shù)據(jù)。import pandas as pdimport numpy as npdate_rng = pd.date_range(start='7/22/2019', end='7/22/2020', freq='S')df = pd.DataFrame(date_rng, columns=['start_timestamp'])df['end_timestamp'] = date_rngdf['start_timestamp'] = np.random.randint(1563753600,1595376000,size=(len(date_rng)))df['end_timestamp'] = np.random.randint(1563753600,1595376000,size=(len(date_rng)))df = df[(df.end_timestamp/df.start_timestamp<=1.0000009)&(df.end_timestamp/df.start_timestamp>=1.0000001)]df = df.to_numpy()df = df.reshape(df.shape[0]*2,1)df = np.sort(df,axis=0)df = df.reshape(int(df.shape[0]/2),2)df = pd.DataFrame(df,columns=['start_timestamp','end_timestamp'])df['start_timestamp'] = pd.to_datetime(df['start_timestamp'],unit='s')df['end_timestamp'] = pd.to_datetime(df['end_timestamp'],unit='s')
1 回答
揚(yáng)帆大魚
TA貢獻(xiàn)1799條經(jīng)驗(yàn) 獲得超9個贊
編寫您的邏輯代碼，一切都很好
freq='S'沒有任何意義，您將生成與開始日期和結(jié)束日期之間的秒數(shù)一樣多的行在隨機(jī)化開始時間后，使用當(dāng)前行和下一行作為結(jié)束時間隨機(jī)函數(shù)的種子。這是作為列表理解嗎
在范圍的開始和結(jié)束處獲取 UTC 秒數(shù)時更聰明一些
import pandas as pd
import numpy as np
from datetime import datetime
# date_rng = pd.date_range(start='5/18/2019', end='7/22/2020', freq='S')
date_rng = pd.date_range(start='5/18/2019', end='5/19/2019', freq='min')
sec = [(date_rng.min() - datetime(1970, 1, 1)).total_seconds(),
(date_rng.max() - datetime(1970, 1, 1)).total_seconds() ]
df = pd.DataFrame(date_rng, columns=['start_timestamp'])
df['start_timestamp'] = np.random.randint(sec[0],sec[1],size=(len(date_rng)))
df = df.sort_values(by="start_timestamp")
l = df["start_timestamp"].tolist() # get randomised start times
l[-1] = sec[1] # set last time to end of range
# randomise end time between two start times
df['end_timestamp'] = [np.random.randint(l[i], l[i+1]) if i<len(l)-1 and l[i]<l[i+1] else l[i] for i, s in enumerate(l)]
df['start_timestamp'] = pd.to_datetime(df['start_timestamp'],unit='s')
df['end_timestamp'] = pd.to_datetime(df['end_timestamp'],unit='s')
添加回答
舉報(bào)
0/150
提交
取消