Jung-Pu-Chang / Member_Classification

Caculate LRFM、NES、CAI

Member_Classification

Using pandas to caculate LRFM NES CAI in python.
Author: Jung-Pu-Chang、容噗玩Data
Date: 2023-08-23
Chinese Explain

Directory

.
├── README.md
├── LICENSE
├── requirements.txt
├── MemberClass.py
└── example.py

Example

import pandas as pd
from MemberClass import MemberClass

data = {
        'ID': ['666', '666', 123, '666', ''],  # any type can be caculate
         # only date with ('%Y-%m-%d','%Y/%m/%d','%d-%m-%Y','%d/%m/%Y') can be caculate
        'DATE': ['2016-01-01', '2016-02-01','2016-12-31','2016-08-01','2016-07-02'],
        'count': [10,20.1,0,'3','4'], # only str, int, float can be caculate
        'cost': [10,10.26,6,'80','100'], # only str, int, float can be caculate
        }
df = pd.DataFrame(data)

test = MemberClass(data = df, memberID = 'ID', date = 'DATE', quantity = 'count', spend = 'cost') 
LRFM = test.LRFM(cur_year = 2017, cur_month = 1, cur_day = 1)
NES = test.NES(cur_year = 2017, cur_month = 1, cur_day = 1, period_days = 366, new_def = 60)
CAI = test.CAI()

Module Description : MemberClass.py

0. Def Contents

def	purpose
init	Transform format
LRFM	Caculate LRFM in python
NES	Caculate NES in python
CAI	Caculate CAI in python

1. def__init__ :

Parameter	Discription	Type
data	raw data	data.frame
memberID	col name of customerID	str
date	col name of date	str
quantity	col name of consuming times	str
spend	col name of consuming amount	str

2. def LRFM :

• L : L_raw_days = first_consuming_date - last_consuming_date, L_raw_days ≥ mean(L_raw_days) = 1, oterwise = 0, every member contact days
• R : R_raw_days = current_date - last_consuming_date, R_raw_days < mean(R_raw_days) = 1, oterwise = 0, how long every member days
• F : count(quantity) ≥ mean(count(quantity)) = 1, oterwise = 0, every member consumes frequency per day (eg : Member A consumes five times(five envoices) in 2023/8/24, his F = 1)
• M : sum(spend) ≥ mean(sum(spend)) = 1, oterwise = 0, Sum of every member amount of consuming
• LRFM_score_sum ∈ [0,4] = L + R + F + M

Parameter	Discription	Type
cur_year	current year	int
cur_month	current month	int
cur_day	current day	int

3. def NES :

• NES : N = new customer = R_raw_days < new_def，E = freq_ratio ≤ 2、2 < S1 ≤ 2.5、2.5 < S2 ≤ 3、3 < S3
• count : every member consumes frequency per day (eg : Member A consumes five times(five envoices) in 2023/8/24, his count = 1)
• freq = count / period_days
• freq_ratio = R_raw_days / freq
• R_raw_days = current_date - last_consuming_date

Parameter	Discription	Type
cur_year	current year	int
cur_month	current month	int
cur_day	current day	int
period_days	raw data period days	int
new_def	new customer definition : current_date - last_consuming_date	int

4. def CAI :

• CAI : < 0 = 0 > 0
• weight :
• avg_freq :
• w_avg_freq :

Refference

Chinese Refference

LRFM
NES
CAI