sqlite条件下随机生成数据

Question

作为创建机器学习练习的一部分，我有一个如下所示的 sqlite 数据库： “Pizza”、“DateDimension”和“Customer”表已填充，有 5 种不同的比萨饼和 10 位顾客。

如何为 sales 表生成随机数据，但给出创建模式的条件？

例如，在星期一/星期二，随机销售数字在 15 到 25 之间，在一周的其他日子随机销售数字在 10 到 20 之间，周末在 20 到 30 之间。

知道一天内，对于 ID=1 的客户销售 15 个比萨饼，ID=2 的客户销售了 2 个 ID=2 的比萨饼和 4 个 ID=3 的比萨饼， 对于 ID=2 的客户，销售了 3 个 ID=2 的比萨饼，并且
对于 ID=3 的客户，销售 3 个 ID=4 的比萨饼和 3 个 ID=2 的比萨饼。

这样一共有 15 个比萨饼，其中 8 个 ID=2 比萨饼，4 个 ID=3 比萨饼，3 个 ID=4 个比萨饼，分给三个顾客。

我不知道我的解释是否足够清楚，基本上我们每天需要一个随机的比萨销售数量，并且这个数字本身是在比萨类型和顾客之间随机分配的，在某些情况下条件，以便遵循每天销售数量的模式。

Answer 1

我相信以下内容可能符合您的要求。

首先，它在日期范围（见 cmets）中每 10 分钟创建一次销售额，到中间表 (initial_sales) 中将有大约 10k 左右。每个人都会有一个随机的披萨、客户和数量。

下一个中间表 daily_sales 是从按日期分组的 initial_sales 生成的（即每个日期 1 行），除了日期是星期几（0 = Sun ..... 6= Sat）（不是真的需要）更重要的是第三列，它有一个基于星期几的随机数（CASE WHEN THEN END 结构）。

然后查询 daily_sales 表，并为每一行生成 sql 以将数据插入 pizzeriasales 表。例如

INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-01' ORDER BY random() LIMIT 13;
INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-02' ORDER BY random() LIMIT 25;
INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-03' ORDER BY random() LIMIT 29;
INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-04' ORDER BY random() LIMIT 20;
INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-05' ORDER BY random() LIMIT 17;
INSERT INTO pizzeriasales SELECT null,date(date),time(date),pizzaId,quantity,clientid FROM initial_sales WHERE date(date) = '2021-01-06' ORDER BY random() LIMIT 10;

• 每个人都会从相关日期的initial_sales中选择随机行（ORDER BY random()），根据daily_sales中的数字选择行数 表。该数字是星期几的较小数字与较小数字加 10 之间的随机数。

作为一个例子，它来自按日期分组时生成的 pizzeriasales 表，第一列显示当天的订单数。

(12, 1, '2021-01-01', '16:10:00', 5, 3, 8)
(23, 13, '2021-01-02', '01:40:00', 3, 3, 1)
(28, 36, '2021-01-03', '05:00:00', 3, 1, 5)
(16, 64, '2021-01-04', '10:10:00', 5, 2, 9)
(22, 80, '2021-01-05', '09:10:00', 4, 2, 7)
(14, 102, '2021-01-06', '10:10:00', 1, 1, 3)
(20, 116, '2021-01-07', '05:50:00', 5, 2, 4)

• 请注意，除了第一个值之外，重要的值 12 表示 2021 年 1 月 1 日的销售额是多少（看起来是星期五，因为 23 日和 28 日接下来的两个日期在 20- 30 范围）。

• 注意您可能会注意到，只要日期采用 SQLite 可识别的格式 (https://sqlite.org/lang_datefunc.html)，就不需要 DateDimension 表

Pyhton 明智的做法是：-

import sqlite3
conn = sqlite3.connect("pizzasales.db")
cur = conn.cursor()

# Drop TABLES for clean TESTING environment
cur.execute("DROP TABLE IF EXISTS pizzeriasales;")
cur.execute("DROP TABLE IF EXISTS pizza")
cur.execute("DROP TABLE IF EXISTS client")
# Create and populate the pizza table
cur.execute("CREATE TABLE IF NOT EXISTS pizza (id INTEGER PRIMARY KEY, name TEXT, price REAL);")
cur.execute("INSERT INTO pizza VALUES "
            "(null,'Meat Lovers', 10.00)"
            ",(null,'Hawian',12.50)"
            ",(null,'Neopolitan',11.00)"
            ",(null,'Margerita',9.00),(null,'standard',8.00);"
            )
# Create and populate the clieant table
cur.execute("CREATE TABLE IF NOT EXISTS client (id INTEGER PRIMARY KEY, name TEXT, firstName TEXT);")
cur.execute("INSERT INTO client VALUES "
            "(null,'Mary','Smith')"
            ",(null,'Fred','Bloggs')"
            ",(null,'Sarah','Bull')"
            ",(null,'Alan','Taylor')"
            ",(null,'Linda','James')"
            ",(null,'Barbara','Astley')"
            ",(null,'Mark','Jones')"
            ",(null,'Catherine','Seymour')"
            ",(null,'Charles','Page')"
            ",(null,'Alex','Howard');"
            )
# Create the pizzeriasales table to be populated later
cur.execute("CREATE TABLE IF NOT EXISTS pizzeriasales (id INTEGER PRIMARY KEY, date TEXT, hour TEXT, pizzaId INTEGER, quantity INTEGER, clientID INTEGER);")

# Create the intermediate initial sales (individual orders every 10 minutes with random pizza and client)
cur.execute("DROP TABLE IF EXISTS initial_sales;")
cur.execute("CREATE TABLE IF NOT EXISTS initial_sales (date TEXT,pizzaId INTEGER, clientId INTEGER, quantity INTEGER);")
# Populate the table (some 10000 rows)
cur.execute("WITH "
            "cte_sales(date,pizzaId,clientId,quantity) "
            " AS (SELECT '2021-01-01 11:00'" # <<<<<<< date to start from
            ",abs(random() % (SELECT max(id) FROM pizza)) + 1" # <<<<<<<< random pizza
            ",abs(random() % (SELECT max(id) FROM client)) + 1" # <<<<<<<< random client
            ",abs(random() % 3) + 1" # <<<<<<<< random quantity
            " UNION ALL "
            " SELECT "
            "datetime(date,'+10 minute')"
            ", (1 + abs(random() % (SELECT max(id) FROM pizza)))"
            ",(1 + abs(random() % (SELECT max(id) FROM client)))"
            ",abs(random() % 3) + 1 FROM cte_sales "
            "WHERE date < '2021-03-01' " # <<<<<<< date to end on
            "LIMIT 10000" # <<<<<<<< secondary limit just in case
            ") "
            "INSERT INTO initial_sales SELECT * FROM cte_sales;"
            )

# Create the base dates i.e. 1 date with the number of sales that day
cur.execute("DROP TABLE IF EXISTS daily_sales_numbers;")
cur.execute("CREATE TABLE IF NOT EXISTS daily_sales_numbers (date TEXT, day_in_week,number);")
# Now populate the daily_sales_numbers
cur.execute("WITH "
            "cte_base_dates(date,day_in_week,number) AS ("
            "SELECT "
            "date, "
            "strftime('%w',date),"
            "CASE "
            "WHEN CAST(strftime('%w',date) AS INTEGER) BETWEEN 1 AND 2 THEN 15 + abs(random() % 11) "
            "WHEN CAST(strftime('%w',date) AS INTEGER) BETWEEN 3 AND 5 THEN 10 + abs(random() % 11) "
            "ELSE 20 + abs(random() % 11) END "
            "FROM initial_sales "
            "GROUP BY date(date)"
            ") "
            "INSERT INTO daily_sales_numbers SELECT * FROM cte_base_dates;")

# Now prepare to build the pizzeriasales table
cursor = cur.execute("SELECT * FROM daily_sales_numbers;")
sql = []
for row in cursor:
    sql.append("INSERT INTO pizzeriasales " \
          "SELECT null,date(date),time(date),pizzaId,quantity,clientid " \
          "FROM initial_sales " \
          "WHERE "
               "date(date) = '" + (row[0])[0:10] #<<<<<<<< date as per the respective daily_sales row
               + "' ORDER BY random() " #<<<<<<<< random selection from the initial_sales
                 "LIMIT " + str(row[2]) #<<<<<<<< the number to include based upon the day of the week
               + ";")
for stmnt in sql:
    print(stmnt)
    cur.execute(stmnt)

# ALL done so show the number of rows in pizzeriasales and a per date basis
# note other data is from arbritary rows of the group (i.e. date)
conn.commit()
cursor = cur.execute("SELECT count(*),* FROM pizzeriasales GROUP BY date(date)")
for row in cursor:
    print(row)

cursor.close()
conn.commit()
cur.close()
conn.close()