淘宝数据,数据处理,时间序列分析,获客分析,购买路径分析

-- 创建数据库
create database taobao;
use taobao;
desc use_behaviour;-- 创建表格
create table use_behaviour(
user_id int(9),
item_id int(9),
category_id int(9),
behaviour_type varchar(5),
timestamps  int(14));-- 查询已导入多少条
select  count(*)  from use_behaviour;
select  *  from use_behaviour limit 10;#改变字段名timestamp 改成timestamps
alter table use_behaviour change  timestamp timestamps int(14);
desc use_behaviour;-- -检查空值
-- select * from   where   ziduanming  is  null-- 检查重复值
select user_id, item_id, timestamps from use_behaviour
group by user_id,item_id,timestamps
having count(*) > 1;-- -去重
alter table use_behaviour add id  int first;
select  *  from use_behaviour limit 10;
-- 将id设置成自增主键
alter table   use_behaviour modify id int primary key  auto_increment;
select  *  from use_behaviour limit 10;delete use_behaviour from 
use_behaviour, 
(
select user_id,item_id, timestamps, min(id) id 
from use_behaviour
group by user_id,item_id,timestamps
having count(*) > 1
) t2
where use_behaviour.user_id=t2.user_id
and use_behaviour.item_id=t2.item_id
and use_behaviour.timestamps=t2.timestamps 
and use_behaviour.id>t2.id;-- 增加三个字段 date time hour
-- 更改buffer值
show variables like '%_buffer%';set global  innodb_buffer_pool_size = 10700000000;alter table use_behaviour  add datetimes TIMESTAMP(0);
update use_behaviour set  datetimes = FROM_UNIXTIME(timestamps);
select * from use_behaviour limit 5;-- date
alter table use_behaviour  add dates char(10);
alter table use_behaviour  add times char(8);
alter table use_behaviour  add hours char(2);
-- 一次性对三个字段进行截取活分开截取,分开截取将多花两倍时间
update use_behaviour set dates=substring(datetimes,1,10 ),times=substring(datetimes,12,8 ),dates=substring(datetimes,12,2);update use_behaviour set dates=substring(datetimes,1,10 );
update use_behaviour set times=substring(datetimes,12,8 );
update use_behaviour set hours=substring(datetimes,12,2);
select * from use_behaviour limit 5;-- 去异常;三部曲:去空去重去异常
select max(datetimes),min(datetimes) from use_behaviour;delete from use_behaviour
where datetimes <'2017-11-25 00:00:00'
or datetimes > '2017-12-03 23:59:59'
-- 共删除942行-- 数据概览
desc use_behaviour;
select * from use_behaviour limit 5;
select count(1) from use_behaviour; -- 1889658条记录-- 创建临时表
drop  table if exists temp_behaviour;
create table temp_behaviour  like use_behaviour;-- 截取
insert into temp_behaviour
select * from use_behaviour limit 100000;select *  from  temp_behaviour limit  5;-- pv
select dates
, count(*) 'pv'
from temp_behaviour 
where behaviour_type ='pv'
group by dates;-- 独立访客数uv
select dates
, count(distinct user_id) 'uv'
from temp_behaviour 
where behaviour_type ='pv'
group by dates;-- 一条语句
select dates
, count(*) 'pv'
, count(distinct user_id) 'uv'
,round(count(*)/count(distinct user_id),1) 'pv/uv'
from temp_behaviour 
where behaviour_type ='pv'
group by dates;-- 处理真实数据
create table pv_uv_puv(
dates  char(10),
pv int(9),
uv int(9),
puv decimal(10,1)
);insert into pv_uv_puv
select dates
, count(*) 'pv'
, count(distinct user_id) 'uv'
,round(count(*)/count(distinct user_id),1) 'pv/uv'
from use_behaviour 
where behaviour_type ='pv'
group by dates;-- 测试
select * from pv_uv_puv-- 去除异常数据
delete  from  pv_uv_puv  where dates is null;
delete from use_behaviour where dates is null;-- 留存率
select  user_id,dates
from  temp_behaviour
group by user_id,dates;-- 自关联  相同的userid以及b的日期比a的日期大的数据
select * from 
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) a
,
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) b
where a.user_id = b.user_id
and a.dates<b.dates;-- 次日留存数,即日期相隔一天的
select a.dates
,count(if (datediff(b.dates,a.dates)=0, b.user_id, null)) retention_0
,count(if (datediff(b.dates,a.dates)=1, b.user_id, null)) rentention_1
,count(if (datediff(b.dates,a.dates)=3, b.user_id, null)) rentention_33
from 
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) a
,
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) b
where a.user_id = b.user_id
and a.dates<=b.dates
group by a.dates;-- 留存率
select a.dates
,count(if (datediff(b.dates,a.dates)=1, b.user_id, null))/count(if (datediff(b.dates,a.dates)=0, b.user_id, null))  rentention_1from 
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) a
,
(
select  user_id,dates
from  temp_behaviour
group by user_id,dates
) b
where a.user_id = b.user_id
and a.dates<=b.dates
group by a.dates;-- 保存结果
create table retention_rate(
dates  char(10),
retention_1 float
)insert into retention_rate
select a.dates
,count(if (datediff(b.dates,a.dates)=1, b.user_id, null))/count(if (datediff(b.dates,a.dates)=0, b.user_id, null))  rentention_1from 
(
select  user_id,dates
from  use_behaviour
group by user_id,dates
) a
,
(
select  user_id,dates
from  use_behaviour
group by user_id,dates
) b
where a.user_id = b.user_id
and a.dates<=b.dates
group by a.dates;-- 跳失率
-- 跳使用户
select count(*)
from
(
select user_id from use_behaviour
group by user_id
having count(behaviour_type)=1
) aselect  sum(pv) from pv_uv_puv; -- 1782280-- 1/1782280-- 时间序列分析
select dates,hours
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type='buy',behaviour_type,null)) 'buy'
from temp_behaviour 
group by dates,hours
order by dates,hours--存储
create  table date_hour_behaviour(
dates char(10),
hours char(2),
pv int,
cart int,
fav int,
buy int
);--结果插入
insert into date_hour_behaviour
select dates,hours
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type='buy',behaviour_type,null)) 'buy'
from use_behaviour 
group by dates,hours
order by dates,hoursselect * from date_hour_behaviour-- 统计各类行为数
select behaviour_type
,count(DISTINCT user_id) user_num
from temp_behaviour
group by behaviour_type
order by behaviour_type  desc;-- 存储 ARCHAR类型用于存储可变长度字符串 CHAR类型用于存储固定长度字符串
create table behaviour_user_num(
behaviour_type varchar(5),
user_num int
);insert into behaviour_user_num
select behaviour_type
,count(DISTINCT user_id) user_num
from use_behaviour
group by behaviour_type
order by behaviour_type  desc;
-- 测试
select *  from behaviour_user_num;-- 转化率分析  0.6844
select 12630/18453-- 统计各类行为数量
select behaviour_type
,count(*) user_num
from temp_behaviour
group by behaviour_type
order by behaviour_type  desc;-- 存储各类行为数据数量
create table behaviour_num(
behaviour_type varchar(5),
behaviour_num int
);insert into behaviour_num
select behaviour_type
,count(*) behaviour_count_num
from use_behaviour
group by behaviour_type
order by behaviour_type  desc;
-- 检查是否成功
select  * from  behaviour_num;-- 行为路径分析select  user_id,item_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
from temp_behaviour
group by user_id,item_idcreate table use_behaviout_view(
uesr_id int(9),
item_id int(9),
pv int,
cart int,
fav int,
buy int
);insert into use_behaviout_view
select  user_id,item_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
from use_behaviour
group by user_id,item_id-- 修改字段名
alter table use_behaviout_view change uesr_id user_id int;
select * from use_behaviout_view-- 修改表名
ALTER  TABLE use_behaviout_view RENAME TO use_behaviour_view
select * from use_behaviour_view--  用户行为标准化
create view user_behaviour_standard as
select user_id
,item_id
,(case when pv>0 then 1 else 0 end) 浏览了
,(case when fav>0 then 1 else 0 end) 收藏了
,(case when cart>0 then 1 else 0 end) 加购了
,(case when buy>0 then 1 else 0 end) 购买了
from use_behaviour_view-- 路径类型select *
,concat(浏览了,收藏了,加购了,购买了) 购买路径类型
from user_behaviour_standard as a
where a.购买了>0--统计各类购买行为数量
create view path_count as
select 购买路径类型
,count(*) 数量
from 
(
select *
,concat(浏览了,收藏了,加购了,购买了) 购买路径类型
from user_behaviour_standard as a
where a.购买了>0
) b
group by 购买路径类型
order by 数量  desc;create table renhua(
path_type char(4),
description varchar(40)
);insert into  renhua
values('0001','直接购买了'),
('1001','浏览后购买了'),
('0011','加购后购买了'),
('1011','浏览加购后购买了'),
('0101','收藏后购买了'),
('1101','浏览收藏后购买了'),
('0111','收藏加购后购买了'),
('1111','浏览收藏加购后购买了')select  * from renhuaselect * from path_count p
join renhua r
on p.购买路径类型 = r.path_type
order by 数量 desc;-- 存储
create table path_result(
path_type char(4),
description varchar(40),
num int
);insert into path_result
select path_type, description, 数量 num
from
path_count p
join renhua r
on p.购买路径类型 = r.path_type
order by 数量 desc;select * from path_resultselect sum(buy)
from use_behaviour_view 
where buy>0 and fav=0 and cart=0
-- 28790-- 用户定位1 付费和非付费 2RFM模型: 指标(R值:最近一次消费  F值:消费频率  M值:消费金额)+根据F值和R值分类(价值用户、发展用户、保持用户、挽留用户)-- 最近购买时间
select user_id
,max(dates) '最近购买时间'
from temp_behaviour
where behaviour_type='buy'
group by user_id
order by 2 desc;
-- ASC表示按升序排序,DESC表示按降序排序  2 表示按照第二栏-- 购买次数
select user_id
,count(user_id) '购买次数'
from temp_behaviour
where behaviour_type='buy'
group by user_id
order by 2 desc;-- 统一select user_id
,max(dates) '最近购买时间'
,count(user_id) '购买次数'
from temp_behaviour
where behaviour_type='buy'
group by user_id
order by 2 desc, 3 desc; -- 优先按照第二列排序,二裂一样的按照第三列排序drop table if exists rfm_model;
create table rfm_model(
user_id int,
frequency int,
recent char(10)
)insert into rfm_model
select user_id
,count(user_id) '购买次数'
,max(dates) '最近购买时间'
from use_behaviour
where behaviour_type='buy'
group by user_id
order by 2 desc, 3 desc; -- 优先按照第二列排序,二裂一样的按照第三列排序select * from rfm_model;-- 根据最近购买时间对用户进行分层
alter table rfm_model add column rscore int;update rfm_model
set rscore = case
when recent ='2017-12-03' then 5
when recent in ('2017-12-01','2017-12-02') then 4
when recent in ('2017-11-29','2017-11-28') then 3
when recent in ('2017-11-27','2017-11-26') then 2
else 1
end-- 根据购买频次对用户进行分层
alter table rfm_model add column fscore int;select max(frequency),min(frequency) from  rfm_model; -- 72 1update rfm_model
set fscore = case
when frequency between 72 and 40 then 5
when frequency between 21 and 39 then 4
when frequency between 11 and 200 then 3
when frequency between 5 and 10 then 2
else 1
end--  分层
set @f_avg=null;
set @r_avg=null;
select avg(fscore) into @f_avg from rfm_model;
select avg(rscore) into @r_avg from rfm_model;select * 
,(case 
when fscore>@f_avg and rscore>@r_avg then '价值用户'
when fscore>@f_avg and rscore<@r_avg then '保持用户'
when fscore<@f_avg and rscore>@r_avg then '发展用户'
when fscore<@f_avg and rscore<@r_avg then '挽留用户'
end) class
from rfm_model--  将结果插入alter table  rfm_model add column class varchar(40);
update rfm_model
set class =case
when fscore>@f_avg and rscore>@r_avg then '价值用户'
when fscore>@f_avg and rscore<@r_avg then '保持用户'
when fscore<@f_avg and rscore>@r_avg then '发展用户'
when fscore<@f_avg and rscore<@r_avg then '挽留用户'
end;select * from rfm_model limit 10;--  统计各分区用户数select class,count(distinct user_id)  from rfm_model
group by class--  商品按照热度分类select  category_id
,count(if(behaviour_type='pv',behaviour_type,null)) '品类浏览量'
from temp_behaviour 
group by category_id
order by 2 desc
limit 10;select  item_id
,count(if(behaviour_type='pv',behaviour_type,null)) '商品浏览量'
from temp_behaviour 
group by item_id
order by 2 desc
limit 10;select category_id,item_id,品类商品浏览量
from 
(
select  category_id,item_id
,count(if(behaviour_type='pv',behaviour_type,null)) '品类商品浏览量'
,rank() over(partition by category_id order by count(if(behaviour_type='pv',behaviour_type,null))  desc) r
from temp_behaviour 
group by category_id,item_id
order by 3 desc
) a
where r =1 
order by a.品类商品浏览量 desc
limit 10create table popular_categories(
category_id int,
pv int);create table popular_items(
item_id int,
pv int);create table popular_cateitems(
category_id int,
item_id int,
pv int);insert into popular_categories
select  category_id
,count(if(behaviour_type='pv',behaviour_type,null)) '品类浏览量'
from use_behaviour 
group by category_id
order by 2 desc
limit 10;insert into popular_items
select  item_id
,count(if(behaviour_type='pv',behaviour_type,null)) '品类浏览量'
from use_behaviour 
group by item_id
order by 2 desc
limit 10;insert into popular_cateitems
select  category_id
,item_id
,count(if(behaviour_type='pv',behaviour_type,null)) '品类浏览量'
from use_behaviour 
group by category_id,item_id
order by 3 desc
limit 10;select * from popular_cateitems;-- 特定商品转化率
select  item_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
,count(distinct if(behaviour_type='buy', user_id,null))/count(distinct user_id) 商品转化率
from use_behaviour
group by item_id
order by 商品转化率 desc;--  保存
create table item_detail(
item_id int,
pv int,
cart int,
fav int,
buy int,
user_buy_rate float
);insert into item_detail
select  item_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
,count(distinct if(behaviour_type='buy', user_id,null))/count(distinct user_id) 商品转化率
from use_behaviour
group by item_id
order by 商品转化率 desc;select  * from item_detail-- 品类转化率
select  category_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
,count(distinct if(behaviour_type='buy', user_id,null))/count(distinct user_id) 品类转化率
from use_behaviour
group by category_id
order by 品类转化率 desc;--  保存
create table category_detail(
category_id int,
pv int,
cart int,
fav int,
buy int,
user_buy_rate float
);insert into category_detail
select  category_id
,count(if(behaviour_type ='pv',behaviour_type,null)) 'pv'
,count(if(behaviour_type ='cart',behaviour_type,null)) 'cart'
,count(if(behaviour_type ='fav',behaviour_type,null)) 'fav'
,count(if(behaviour_type ='buy',behaviour_type,null)) 'buy'
,count(distinct if(behaviour_type='buy', user_id,null))/count(distinct user_id) 品类转化率
from use_behaviour
group by category_id
order by 品类转化率 desc;select  * from category_detail-- 商品特征分析  tableau  添加平均值参考线  筛选器-- 数据可视化

在这里插入图片描述
在这里插入图片描述
在这里插入图片描述
在这里插入图片描述

在这里插入图片描述

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处:http://www.mzph.cn/news/477986.shtml

如若内容造成侵权/违法违规/事实不符,请联系多彩编程网进行投诉反馈email:809451989@qq.com,一经查实,立即删除!

相关文章

谁说发 paper 一定要追快打新?2021年,研究 word2vec 也能中顶会!

文 | jxyxiangyu前言“小夕&#xff0c;小夕&#xff0c;你关注的任务sota又被刷新了&#xff01;”“什么&#xff1f;&#xff01;”还在跑实验的小夕默默流下了辛酸泪不得不说nlp领域的发展真的太快了&#xff0c;炼丹师们不光要时刻关注前沿热点&#xff0c;还要快速做出实…

论文浅尝 | Multilingual LAMA: 探索多语言预训练语言模型中的知识

笔记整理&#xff1a;谭亦鸣&#xff0c;东南大学博士生来源&#xff1a;EACL‘21链接&#xff1a;https://aclanthology.org/2021.eacl-main.284.pdf概述本文关注将语言模型&#xff08;LM&#xff09;视作一个知识库&#xff0c;然后用于解决例如句子填空这样的NLP任务&#…

LeetCode 860. 柠檬水找零(贪心)

1. 题目 在柠檬水摊上&#xff0c;每一杯柠檬水的售价为 5 美元。 顾客排队购买你的产品&#xff0c;&#xff08;按账单 bills 支付的顺序&#xff09;一次购买一杯。 每位顾客只买一杯柠檬水&#xff0c;然后向你付 5 美元、10 美元或 20 美元。你必须给每个顾客正确找零&…

召回粗排精排-级联漏斗(下)

文 | 水哥源 | 知乎saying召回区分主路和旁路&#xff0c;主路的作用是个性化向上管理&#xff0c;而旁路的作用是查缺补漏。推荐系统的前几个操作可能就决定了整个系统的走向&#xff0c;在初期一定要三思而后行。做自媒体&#xff0c;打广告&#xff0c;漏斗的入口有多大很重…

学术联赛 | 运用知识图谱技术,赋能多领域应用 ——“未来杯”AI学术联赛总决赛暨颁奖典礼圆满落幕...

本文转载自公众号&#xff1a;AI学习社。由北京大学软件工程国家工程研究中心主办&#xff0c;华为终端有限公司及中软国际教育科技集团全程战略支持&#xff0c;STEER TECH科技平台、北京乐智元素科技有限公司、艾肯文化传媒&#xff08;北京&#xff09;有限公司、AI TIME承办…

LeetCode 1013. 将数组分成和相等的三个部分

1. 题目 给定一个整数数组 A&#xff0c;只有我们可以将其划分为三个和相等的非空部分时才返回 true&#xff0c;否则返回 false。 形式上&#xff0c;如果我们可以找出索引 i1 < j 且满足 (A[0] A[1] … A[i] A[i1] A[i2] … A[j-1] A[j] A[j-1] … A[A.lengt…

谷歌 | 多任务学习,如何挑选有效的辅助任务?只需一个公式!

文 | 小伟编 | 小轶前言说到多任务学习&#xff0c;大家都不陌生&#xff0c;不管是在学术界还是工业界都已经有了很多成熟的探索与应用。在理想的多任务学习中&#xff0c;各个任务对彼此应当是有益的&#xff0c;所有任务相互促进&#xff0c;从而达到超过单任务学习的效果。…

LeetCode 888. 公平的糖果交换(哈希set)

文章目录1. 题目2. 解题2.1 暴力查找2.2 哈希set1. 题目 爱丽丝和鲍勃有不同大小的糖果棒&#xff1a;A[i] 是爱丽丝拥有的第 i 块糖的大小&#xff0c;B[j] 是鲍勃拥有的第 j 块糖的大小。 因为他们是朋友&#xff0c;所以他们想交换一个糖果棒&#xff0c;这样交换后&#…

OpenKG开源系列 | 面向知识的推理问答编程语言KoPL(清华大学)

OpenKG地址&#xff1a;http://openkg.cn/tool/koplGitHub地址&#xff1a;https://github.com/THU-KEG/KoPL网站地址&#xff1a;https://kopl.xlore.cn开放许可协议&#xff1a;MIT License贡献者&#xff1a;清华大学&#xff08;曹书林、史佳欣、姚子俊、吕鑫、聂麟骁、逄凡…

前端组件化开发实践

前言 一位计算机前辈曾说过&#xff1a; Controlling complexity is the essence of computer programming.随着前端开发复杂度的日益提升&#xff0c;组件化开发应运而生&#xff0c;并随着 FIS、React 等优秀框架的出现遍地开花。这一过程同样发生在美团&#xff0c;面临业务…

LeetCode 937. 重新排列日志文件(自定义排序)

1. 题目 你有一个日志数组 logs。每条日志都是以空格分隔的字串。 对于每条日志&#xff0c;其第一个字为字母数字标识符。然后&#xff0c;要么&#xff1a; 标识符后面的每个字将仅由小写字母组成&#xff0c;或&#xff1b;标识符后面的每个字将仅由数字组成。 我们将这…

预训练时代微调新范式,高性能加速2800%,NLPer赶紧看过来!

一、导读PaddleNLP 是兼具科研学习和产业实践能力的 Python NLP 工具包&#xff0c;提供中文领域丰富的预训练模型和部署工具&#xff0c;被高校、企业开发者广泛应用。近日&#xff0c;PaddleNLP v2.1正式发布&#xff0c;为开发者带来三项重要更新&#xff1a;开箱即用的工业…

论文浅尝 | GaussianPath: 用于知识图谱推理的贝叶斯多跳推理框架

笔记整理&#xff1a;谭亦鸣&#xff0c;东南大学博士生来源&#xff1a;AAAI’21链接&#xff1a;https://ojs.aaai.org/index.php/AAAI/article/view/16565多跳推理由于对下游任务例如问答和图谱补全的可解释性受到关注。多跳推理是一个典型的顺序决策过程&#xff0c;可表述…

AllenAI 发布万能问答系统 MACAW!各类题型样样精通,性能大幅超越 GPT-3!

文 | python前言GPT-3 等超大规模预训练语言模型&#xff0c;在少监督任务&#xff08;few-shot tasks&#xff09;上取得了令人瞩目的成绩。而这篇文章中&#xff0c;AllenAI的研究员提出了大规模生成式问答模型&#xff0c;MACAW。基于多角度预训练&#xff0c;MACAW可以用于…

论文浅尝 | SMBOP: Semi-autoregressive Bottom-up Semantic Parsing

笔记整理&#xff1a;陈永锐&#xff0c;东南大学博士来源&#xff1a;NAACL 2021概述近年来语义解析的事实上的标准解码方法是使用自顶向下的深度优先遍历对目标程序的抽象语法树进行自回归解码。该工作提出了一种替代方法&#xff1a;半自回归自底向上解析器&#xff08;SMBO…

美团酒店Node全栈开发实践

前后端分离的背景 “前后端分离”显然已不是什么新鲜的话题&#xff0c;Zakas在2013年10月份就曾发表过一篇博客《Node.js and the new web front-end》讨论Node背景下新时代的前端。毫无疑问&#xff0c;Node的出现给JavaScript语言带来了新的生机&#xff0c;也使得前端开发者…

统计学习方法总结

统计学习方法总结 阅读目录(Content)0. 相关知识点0x1: 监督学习1. 模型假设空间2. 生成模型与判别模型的联系与区别 3. 学习策略4. 分类问题与回归问题5. 利用模型进行预测和分析0x2&#xff1a;模型评估与模型选择1. 训练误差与测试误差2. 过拟合与模型选择0x3&#xff1a;正…

LeetCode 997. 找到小镇的法官(图的出度和入度)

1. 题目 在一个小镇里&#xff0c;按从 1 到 N 标记了 N 个人。传言称&#xff0c;这些人中有一个是小镇上的秘密法官。 如果小镇的法官真的存在&#xff0c;那么&#xff1a; 小镇的法官不相信任何人。每个人&#xff08;除了小镇法官外&#xff09;都信任小镇的法官。只有…

哈工大|NLP数据增强方法?我有15种

文 | rumor源 | 李rumor卷友们好&#xff0c;我是rumor。十一假期过的太快了&#xff0c;不知道你们缓过来没有&#xff0c;没有的话今天我们就来一起读一篇综述缓缓&#xff0c;弥补假期没学习的遗憾。这篇40多页的综述出自哈工大车万翔老师的团队&#xff0c;一共总结了15种N…

论文浅尝 | Wordly Wise(WoW) - 用于语音视觉知识问答的跨语言知识融合模型

笔记整理: 谭亦鸣&#xff0c;东南大学博士生来源&#xff1a;NAACL’21链接&#xff1a;https://aclanthology.org/2021.naacl-main.153.pdf论文提出了一种新的知识图谱问答数据集命名为FVSQA&#xff0c;这是一种语音视觉知识问答类型的任务&#xff0c;即问题形式为音频&…