https://pandas.pydata.org/docs

Import thư viện:

import pandas as pd
import numpy as np
 
pd.options.mode.copy_on_write = True # https://pandas.pydata.org/docs/user_guide/copy_on_write.html

df = pd.read_csv('data/pandas/file.csv')
df = pd.read_csv('data/pandas/file.csv', engine="pyarrow")
df = pd.read_csv('data/pandas/file.csv', low_memory=False)

tempdict = {'col1':[1,2,3], 'col2':[4,5,6], 'col3':[7,8,9]}
 
df = pd.DataFrame.from_dict(tempdict)

df_new = df[["eid", "type", "request_amount", "user_name", "status", "partner_status", "approved_time", "created_at", "updated_at", "date"]]

df_ut_01 = pd.read_csv('../data/ns/ut_2021_01.csv', engine="pyarrow")
df_ut_02 = pd.read_csv('../data/ns/ut_2021_12.csv', engine="pyarrow")
 
df_ut = df_ut_01._append(df_ut_02)

df.head(10)
 
df.head()
 
df.tail(15)

df.columns
 
df.dtypes

df.describe()
 
df.describe(include='object')

df.COLUMN_NAME
 
df['COLUMN_NAME']
 
df[['COLUMN_NAME_1', 'COLUMN_NAME_2']]

df[df['COLUMN_NAME'] == 'Column Value']
 
df[(df['COLUMN_NAME_1']=='Column Value') & (df['COLUMN_NAME_2']==True)]

Ví dụ 1:

approve_from_time = np.datetime64('2021-01-01 00:00:00')
approve_to_time = np.datetime64('2021-01-30 23:59:59')
 
df_ut_data = df_ut[(df_ut['approved_time'] >= approve_from_time) & (df_ut['approved_time'] <= approve_to_time)]
 
df_ut_data_success = df_ut[(df_ut['partner_status'] == 'SUCCESS') & (df_ut['approved_time'] >= approve_from_time) & (df_payout['approved_time'] <= approve_to_time)]

Ví dụ 2:

remit_accounts = ['TerraPay', 'tranglo', 'Nium', 'GME', 'OnePIP', 'Ipay', 'SingX', 'brdgx', 'hanshan', 'HanshanAPI', 'tranglo_CITAD', 'Nium_CITAD', 'SingX_CITAD']
 
df_payout_remit = df_ut_2024_new[
        (df_ut_2024_new['type'] == 'TRANSFER_BANK') 
        & (df_ut_2024_new["user_name"].isin(remit_accounts))
]

df['NEW_COLUMN'] = df['NEW_COLUMN_1'] + df['NEW_COLUMN_2']

df_ut['date'] = pd.to_datetime(df_ut['date'], format='%d/%m/%Y')
df_ut['approved_time'] = pd.to_datetime(df_ut['approved_time'], format='%d/%m/%Y %H:%M:%S')
df_ut['created_at'] = pd.to_datetime(df_ut['created_at'], format='%d/%m/%Y %H:%M:%S')
df_ut['updated_at'] = pd.to_datetime(df_ut['updated_at'], format='%d/%m/%Y %H:%M:%S')
df_ut['processed_time'] = (df_ut['approved_time'] - df_ut['created_at']).dt.seconds
df_ut['weekday'] = df_ut['created_at'].dt.day_name()
df_ut['fee_postpaid'] = np.where(((df_ut['fee_type'] == 'POSTPAID' )) , df_ut['fee'], 0)

df['COLUMN_NAME'] = 'Column Value'

Ví dụ 1:

def get_fee(entity):
    year = int(entity['created_at'].year)
 
    if year == 2024:
        return get_fee_2024(entity)
 
    return 0
 
def get_fee_2024(entity):
    month = int(entity['created_at'].month)
    day = entity['created_at'].day
    provider = entity['provider']
    request_amount = entity['request_amount']    
    bank_no = entity['bank_no']     
 
    if provider == 'EPAY':
        return 2200
 
    if provider == 'TECHCOMBANK' and bank_no != provider:
        return 275 if month >= 3 else 550  
 
    if provider == 'MBBANK':
        return 880 if bank_no != provider else 1980
 
    if provider == 'VIETINBANK' and bank_no != provider:
        return 1650
 
    if provider == 'VPBANK' and bank_no != provider:
        return 275 if month <= 2 else 605
 
    if provider == 'BIDV':
        if bank_no != provider:
            return 1100
 
        return 1705 if request_amount <= 300000000 else 1925  
 
    return 0
 
df['date'] = pd.to_datetime(df['date'], format='%d/%m/%Y')
df['created_at'] = pd.to_datetime(df['created_at'], format='%d/%m/%Y %H:%M:%S')
df['fee'] = df.apply(lambda x: get_fee(x), axis=1)
df['processed_time'] = (df['approved_time'] - df['created_at']).dt.seconds
df.loc[:,'day'] = df.created_at.dt.strftime('%d')

Ví dụ 2:

df_industries = pd.read_csv('../data/ns/user_industries.csv', low_memory=False)
 
df_industries.rename(columns={
    'Tài khoản': 'name', 
    'Tên tài khoản': 'fullname', 
    'Trạng thái': 'status',
    'Thời gian tạo': 'created_at',
    'Ngành nghề': 'industry_name'    
}, inplace=True)
 
df_industries = df_industries.drop(columns=['STT'])
 
df_industries.insert(3, "Note", "")
 
df_industries.set_index('name')
 
df_industries

Ví dụ 3:

import math
 
def get_fee_total(d):
    fee_total = 0
 
    if math.isnan(d['2021_payin_fee_total']) == False:
        fee_total = fee_total + float(d['2021_payin_fee_total'])
    if math.isnan(d['2022_payin_fee_total']) == False:
        fee_total = fee_total + float(d['2022_payin_fee_total'])
    if math.isnan(d['2023_payin_fee_total']) == False:
        fee_total = fee_total + float(d['2023_payin_fee_total'])
    if math.isnan(d['2024_payin_fee_total']) == False:
        fee_total = fee_total + float(d['2024_payin_fee_total'])
 
    if math.isnan(d['2021_payout_fee_total']) == False:
        fee_total = fee_total + float(d['2021_payout_fee_total'])
    if math.isnan(d['2022_payout_fee_total']) == False:
        fee_total = fee_total + float(d['2022_payout_fee_total'])
    if math.isnan(d['2023_payout_fee_total']) == False:
        fee_total = fee_total + float(d['2023_payout_fee_total'])
    if math.isnan(d['2024_payout_fee_total']) == False:
        fee_total = fee_total + float(d['2024_payout_fee_total'])
 
    return fee_total
 
df_fee['fee_total'] = df_fee.apply(lambda x: get_fee_total(x), axis=1)

del df

df_statistic['count'] = df_statistic['count'].fillna(0)

df_statistic['count'] = df_statistic['count'].astype('Int64')

df.groupby(['user_name'])['request_amount'].sum()
 
df_groups = df.groupby(['transaction_id']).agg({'transaction_id':'size', 'fee': 'sum', 'amount': 'mean'}).rename(columns={'transaction_id':'count', 'fee': 'fee_total', 'amount': 'amount_avg'}).reset_index()
 
df_groups.sort_values(by='count', ascending=False)
df_groups.sort_values(by='avg', ascending=False)
 
# Tính giá trị trung bình 1 cột
df["internal_time"].mean()

df_statistic = pd.merge(df_users, df_industries, left_on='name', right_on='name', how = 'left')

df.to_csv('data/pandas/output.csv')
 
df.to_json()
 
df.to_html()
 
with pd.ExcelWriter('../data/ns/dd/file.xlsx') as writer:
    df_1.to_excel(writer, sheet_name='sheet 1')
    df_2_excel(writer, sheet_name='sheet 2')