將數(shù)據(jù)導(dǎo)入hive，通過(guò)SparkSql編寫(xiě)sql,實(shí)現(xiàn)不同業(yè)務(wù)的數(shù)據(jù)計(jì)算實(shí)現(xiàn)，主要講述車(chē)輛卡口轉(zhuǎn)換率，卡口轉(zhuǎn)化率：主要計(jì)算不同卡口下車(chē)輛之間的流向，求出之間的轉(zhuǎn)換率。

在這里插入圖片描述

1、查出每個(gè)地區(qū)下每個(gè)路段下的車(chē)流量

select
	car,
	monitor_id,
	action_time,
	ROW_NUMBER () OVER (PARTITION by car
ORDER by
	action_time) as n1
FROM
	traffic.hive_flow_action

此結(jié)果做為表1，方便后面錯(cuò)位連接使用

在這里插入圖片描述

2、通過(guò)錯(cuò)位連接獲取每輛車(chē)的行車(chē)記錄

通過(guò)表1的結(jié)果，與自身進(jìn)行錯(cuò)位鏈接，并以車(chē)牌為分區(qū)，拼接經(jīng)過(guò)卡口的過(guò)程

(select
	t1.car,
	t1.monitor_id,
	concat(t1.monitor_id,
	"->",
	t2.monitor_id) as way
from
	(
	select
		car,
		monitor_id,
		action_time,
		ROW_NUMBER () OVER (PARTITION by car
	ORDER by
		action_time) as n1
	FROM
		traffic.hive_flow_action) t1
left join (
	select
		car,
		monitor_id,
		action_time,
		ROW_NUMBER () OVER (PARTITION by car
	ORDER by
		action_time) as n1
	FROM
		traffic.hive_flow_action) t2 on
	t1.car = t2.car
	and t1.n1 = t2.n1-1
where
	t2.action_time is not null)

在這里插入圖片描述

獲取到每輛車(chē)的一個(gè)行車(chē)記錄，經(jīng)過(guò)的卡口

3、獲取行車(chē)過(guò)程中的車(chē)輛數(shù)

獲取卡口1~卡口2，…等的車(chē)輛數(shù)有哪些,即拿上面的行車(chē)記錄字段進(jìn)行分區(qū)在進(jìn)行統(tǒng)計(jì)

(select
	s1.way,
	COUNT(1) sumCar
from
	--行車(chē)過(guò)程
(select
		t1.car,
		t1.monitor_id,
		concat(t1.monitor_id,
		"->",
		t2.monitor_id) as way
	from
		(
		select
			car,
			monitor_id,
			action_time,
			ROW_NUMBER () OVER (PARTITION by car
		ORDER by
			action_time) as n1
		FROM
			traffic.hive_flow_action) t1
	left join (
		select
			car,
			monitor_id,
			action_time,
			ROW_NUMBER () OVER (PARTITION by car
		ORDER by
			action_time) as n1
		FROM
			traffic.hive_flow_action) t2 on
		t1.car = t2.car
		and t1.n1 = t2.n1-1
	where
		t2.action_time is not null)s1
group by way)

在這里插入圖片描述

4、獲取每個(gè)卡口的總車(chē)輛數(shù)

獲取每個(gè)卡口最初的車(chē)輛數(shù)，方便后面拿行車(chē)軌跡車(chē)輛數(shù)/總車(chē)輛數(shù)，得出卡口之間的轉(zhuǎn)換率

select
	monitor_id ,
	COUNT(1) sumall
from
	traffic.hive_flow_action
group by
	monitor_id

在這里插入圖片描述

5、求出卡口之間的轉(zhuǎn)換率

select
	s2.way,
	s2.sumCar / s3.sumall zhl
from
	(
	select
		s1.way,
		COUNT(1) sumCar
	from
		--行車(chē)過(guò)程
(
		select
			t1.car,
			t1.monitor_id,
			concat(t1.monitor_id,
			"->",
			t2.monitor_id) as way
		from
			(
			select
				car,
				monitor_id,
				action_time,
				ROW_NUMBER () OVER (PARTITION by car
			ORDER by
				action_time) as n1
			FROM
				traffic.hive_flow_action) t1
		left join (
			select
				car,
				monitor_id,
				action_time,
				ROW_NUMBER () OVER (PARTITION by car
			ORDER by
				action_time) as n1
			FROM
				traffic.hive_flow_action) t2 on
			t1.car = t2.car
			and t1.n1 = t2.n1-1
		where
			t2.action_time is not null)s1
	group by
		way)s2
left join
	--每個(gè)卡口總車(chē)數(shù)
(
	select
		monitor_id ,
		COUNT(1) sumall
	from
		traffic.hive_flow_action
	group by
		monitor_id) s3 on
	split(s2.way,
	"->")[0]= s3.monitor_id

在這里插入圖片描述