基于python、jupyter-notebook 的金融领域用户交易行为分析

说明：本文重在说明交易数据统计、分析方法，所有数据均为生成的数据

　　　　时间原因代码未定义成函数

统计指标：1.用户单日交易行为数据

　　　　　2.按小时为计算单位，统计用户行为数据（旨在求得一天24小时中每个小时的交易金额、交易量和后再做统计计算）

获取数据代码如下：

 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
 __author__ = 'zqf'

 import pymysql
 import random
 import time
 from conf import test_conf

 con = pymysql.connect(**test_conf.con_set)
 print("连接成功")
 cur = con.cursor()
 time_start = time.time()
 for i in range(1, 100000):
     phone_random = random.randint(10000000000, 19999999999)
     transaction_amount_random = random.uniform(0, 50000)
     transaction_type_random = random.randint(1, 5)
     # sql = "select * from my_database.transcation_info"
     sql = "insert into my_database.transaction_info(user_id, name, phone, transaction_amount, transaction_type, " \
           "transaction_time) values('%d', 'name%d', '%d', '%d', '%d', '2019-05-%d %d:%d:%d')" % \
           (random.randint(1, 1000), random.randint(1, 1000), phone_random, transaction_amount_random,
            transaction_type_random, random.randint(1, 31), random.randint(0, 23), random.randint(0, 59),
            random.randint(0, 59))

     cur.execute(sql)
 print("execute完毕即将提交")
 try:
     con.commit()
     print("提交成功")
 except Exception:
     print("插入失败")
     con.rollback()
 time_stop = time.time()
 take_time = time_stop - time_start
 print("花费时间：", take_time)
 # print(cur.fetchall())

在juoyter-notebook中

导入所需包

# 导入所需包
import pandas as pd
import pymysql
from datetime import datetime
import time
from matplotlib import pyplot as plt
plt.rcParams['font.family'] = ['sans-serif']
plt.rcParams['font.sans-serif'] = ['SimHei']
# from conf import test_conf

连接sql数据库

　

 # 连接数据库
 con = pymysql.connect(**{
     'database': 'my_database',
     'host': '192.168.**.**',
     'port': 3306,
     'user': 'root',
     'password': '********',
     'charset': 'utf8'
 })
 print("连接成功")
 cur = con.cursor()
 sql = "select * from transaction_info"
 time_start = time.time()
 df = pd.read_sql(sql=sql, con=con)
 time_stop = time.time()
 take_time = time_stop - time_start
 print("读取十万条数据花费时间：", take_time)
 # print(df)

连接成功
读取十万条数据花费时间： 4.09512186050415

# 用户每天每小时最大交易金额
df_groupby_hour = df.groupby(['user_id', df['transaction_time'].apply(lambda item: datetime.strftime(item, '%Y-%m-%d %H'))])

# 计算每天每小时交易金额最大值， 最小值， 平均值， 交易次数， 交易金额总计
df_calculate_by_hour = df_groupby_hour['transaction_amount'].agg([["daily_max", "max"], ["daily_min", "min"], ["daily_mean", "mean"], ["daily_count", "count"], ["daily_sum", "sum"]])

# 注：每天每小时交易金额最大值为每个小时中累计金额的最大值，所以后续需对hour_sum计算相应参数，hour_max、hour_min、hour_mean的后续相应
# 计算均为以天为单位
df_calculate_by_hour

# 将user_id， transaction_time层索引设置成列索引
df_calculate_by_hour_reset_index = df_calculate_by_hour.reset_index()
df_calculate_by_hour_reset_index

# 再根据user_id、transaction_time 分组，到天
df_calculate_by_hour_reset_index_regroup = df_calculate_by_hour_reset_index.groupby(['user_id', df_calculate_by_hour_reset_index['transaction_time'].apply(lambda item: datetime.strftime(pd.to_datetime(item), '%Y-%m-%d'))])

# 聚合计算
df_finally = df_calculate_by_hour_reset_index_regroup.agg({'daily_max':['max'], 'daily_min':['min'], 'daily_mean':['mean'], 'daily_count':['sum', 'max', 'min', 'median', 'std', 'mean'], 'daily_sum':['sum', 'max', 'min', 'median', 'std', 'mean']})

# 获取用户交易信息
search_user_id = 1
get_user_message = df_finally.loc[search_user_id]

# 将user_id， transaction_time层索引设置成列索引
df_finally_reset = df_finally.reset_index()

df_finally.loc[search_user_id]

# 绘制每天参数图

plt.figure(figsize=(10, 6))
plt.xticks(rotation=45)
x = df_finally.loc[search_user_id].index
li_daily = df_finally.columns.levels[0][:-2]
colors = ['y', 'k', 'r']
for index, i in enumerate(li_daily):
    plt.plot(x, df_finally.loc[search_user_id][i][df_finally[i].columns[0]], label=f"{i}", color=colors[index])

plt.title("每天交易统计")
plt.xlabel("日期")
plt.ylabel("交易金额")
plt.legend()
plt.show()

# 绘制交易量交易图
plt.figure(figsize=(10, 6))
plt.xticks(rotation=45)
daily_count_str = df_finally.columns.levels[0][3]
li_counts = df_finally[daily_count_str].columns
x = df_finally.loc[search_user_id].index
df_finally.columns.levels[0][3]
colors = ['b', 'g', 'c', 'y', 'k', 'r']
for index, i in enumerate(li_counts):
    plt.plot(x, df_finally.loc[search_user_id][daily_count_str][i], label=f"{i}", color=colors[index])

plt.title("每天交易统计")
plt.xlabel("日期")
plt.ylabel("交易次数")
plt.legend()
plt.show()

# 绘制每天每小时交易图
plt.rcParams['font.family'] = ['sans-serif']
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.figure(figsize=(10, 6))
plt.xticks(rotation=45)
x = df_finally.loc[search_user_id].index
li_daily_hour = df_finally.columns.levels[1]
daily_sum_str = df_finally.columns.levels[0][4]
colors = ['b', 'g', 'c', 'y', 'k', 'r']
for index, i in enumerate(li_daily_hour):
    plt.plot(x, df_finally.loc[search_user_id][daily_sum_str][i], label=f"{i}", color=colors[index])

plt.title("每天每小时交易统计")
plt.xlabel("日期")
plt.ylabel("交易金额")
plt.legend()
plt.show()