數據分析之兩種用戶分群方法（RFM和聚類）

本文由於沒有現成的數據，就自己生成了一些商品訂單數據，基於該數據進行了RFM和聚類的構建

1.數據的生成

數據庫表操作

use my_work;

-- 創建商品訂單表
CREATE table goods_orders_ful(
    user_id varchar(100),  -- 用戶id
    order_id varchar(100), -- 訂單id
    is_paid bool, -- 用戶是否實際支付，1支付；0未支付
    amount double, -- 訂單金額
    created_date date, -- 訂單生成日期 yyyy-mm-dd
    created_time timestamp, -- 訂單生成時間 yyyy-mm-dd hh:mm:ss
    business_type varchar(10), -- 業務類型
    region_name varchar(10), -- 所屬區域：如 東部地區
    order_source_name varchar(10), -- 訂單渠道：Web、H5、App 
    is_done bool -- 訂單是否完成
    );
                                            
-- 創建用戶訂單行為中間表                 
drop table if exists user_info_frm_01;
CREATE table user_info_frm_01
as
select gof.user_id,
       sum(gof.amount) all_of_money,
       max(gof.created_date) latest_date, 
       count(gof.order_id) all_of_orders
FROM goods_orders_ful gof 
where gof.is_paid = 1
and gof.is_done = 1
and gof.created_date >= '2020-01-01'
and gof.created_date < '2020-07-01'
group by gof.user_id;

SELECT count(*) from user_info_frm_01 uif;
SELECT * from user_info_frm_01 uif limit 10

-- 創建行為指標均值表
create table if not exists user_info_frm_02
as
select avg(uif.all_of_money) all_of_money_avg,
       avg(datediff('2020-07-22', uif.latest_date)) latest_days_avg,
       avg(uif.all_of_orders) orders_avg
from user_info_frm_01 uif;

SELECT * from user_info_frm_02;
-- 消費均值1107.10，最小天數均值86.9，訂單數量均值2.1

-- 將用戶進行rfm一級打標
create table if not exists user_info_frm_03
as
SELECT uif.user_id,
       case when uif.all_of_money >= 1107.10
            then '高'
            else '低'
            end money,
       case when datediff('2020-07-22', uif.latest_date) >= 86.9
            then '高'
            else '低'
            end recency,
       case when uif.all_of_orders >= 2.1
            then '高'
            else '低'
            end frequency
from user_info_frm_01 uif;

-- 將用戶進行二級打標
create table if not exists user_info_frm_04
as
select uif.user_id,
       uif.recency,
       uif.frequency,
       uif.money,
       case when uif.recency = '高' and uif.frequency = '高' and uif.money = '高'
            then '重要價值用戶'
            when uif.recency = '低' and uif.frequency = '高' and uif.money = '高'
            then '重要保持用戶'
            when uif.recency = '高' and uif.frequency = '低' and uif.money = '高'
            then '重要發展用戶'
            when uif.recency = '低' and uif.frequency = '低' and uif.money = '高'
            then '重要挽留用戶'
            when uif.recency = '高' and uif.frequency = '高' and uif.money = '低'
            then '一般價值用戶'
            when uif.recency = '低' and uif.frequency = '高' and uif.money = '低'
            then '一般保持用戶'
            when uif.recency = '高' and uif.frequency = '低' and uif.money = '低'
            then '一般發展用戶'
            when uif.recency = '低' and uif.frequency = '低' and uif.money = '低'
            then '一般挽留用戶'
            else NULL 
            end type
            
from user_info_frm_03 uif;

python 程序生成數據

 

# _*_ coding: utf-8 _*_ #
# @Time     :2020/7/25 7:30 下午
# @Author   :Zhx


import pymysql
import uuid
import random
import time


class CreateData(object):

    def __init__(self):
        pass

    @staticmethod
    def create():
        user_id_ = random.randint(1, 5000)
        order_id_ = uuid.uuid1()
        is_paid_ = random.choice([1, 0, 1, 1, 1, 1, 1, 1, 1, 1])
        amount_ = random.uniform(10, 1000)
        a1 = (2020, 1, 1, 0, 0, 0, 0, 0, 0)
        a2 = (2020, 6, 31, 23, 59, 59, 0, 0, 0)

        start = time.mktime(a1)  # 生成開始時間戳
        end = time.mktime(a2)  # 生成結束時間戳

        # 隨機生成10個日期字符串
        t = random.randint(start, end)  # 在開始和結束時間戳中隨機取出一個
        date_tuple = time.localtime(t)  # 將時間戳生成時間元組
        created_date_ = time.strftime("%Y-%m-%d", date_tuple)  # 將時間元組轉成格式化字符串
        created_time_ = time.strftime("%Y-%m-%d %H:%M:%S", date_tuple)
        business_type_ = random.randint(0, 20)
        region_name_ = random.choice(['東', '西', '南', '北'])
        order_source_name_ = random.choice(['Web', 'app', 'H5'])
        is_done_ = is_paid_
        return user_id_, order_id_, is_paid_, amount_, created_date_, created_time_, \
            business_type_, region_name_, order_source_name_, is_done_


if __name__ == '__main__':
    database = 'my_work'
    table = 'goods_orders_ful'
    counts = 10000
    create_data = CreateData()
    con = pymysql.connect(database=database, host='localhost',
                          user='root', port=3306, password='199498zhx@')
    cur = con.cursor()
    for i in range(counts):
        user_id, order_id, is_paid, amount, created_date, created_time, \
             business_type, region_name, order_source_name, is_done = create_data.create()
        sql = """insert into %s.%s values('%s', '%s', %d, %f, '%s', '%s', '%s', '%s', '%s', %d)""" % \
              (database, table, user_id, order_id, is_paid, amount, created_date, created_time, business_type,
               region_name, order_source_name, is_done)
        try:
            cur.execute(sql)
            print(i, i % 1000)
            con.commit()
        except Exception as e:
            print(e)
            con.rollback()
    con.close()
    cur.close()

 

源數據字段有：

　　user_id varchar(100),  -- 用戶id

　　order_id varchar(100), -- 訂單id

　　is_paid bool, -- 用戶是否實際支付，1支付；0未支付

      amount double, -- 訂單金額

　　created_date date, -- 訂單生成日期 yyyy-mm-dd

　　created_time timestamp, -- 訂單生成時間 yyyy-mm-dd hh:mm:ss

　　business_type varchar(10), -- 業務類型

　　region_name varchar(10), -- 所屬區域：如 東部地區

　　order_source_name varchar(10), -- 訂單渠道：Web、H5、App 

　　is_done bool -- 訂單是否完成

 

RFM 模型最終表數據

 

 

 

最終的可視化分析使用jupyter完成