实用推荐系统

第1部分　推荐系统的准备工作

第1章　什么是推荐 ..................................................................................3

1.1　现实生活中的推荐 ............................................................................3

1.1.1　推荐系统在互联网上大显身手 .......................................................5

1.1.2　长尾 ..............................................................................................5

1.1.3 Netflix 的推荐系统 ...........................................................................6

1.1.4　推荐系统的定义 .............................................................................13

1.2　推荐系统的分类 .................................................................................15

1.2.1　域 .....................................................................................................16

1.2.2　目的 .................................................................................................16

1.2.3　上下文 .............................................................................................17

1.2.4　个性化级别 .....................................................................................17

1.2.5　专家意见 .........................................................................................19

1.2.6　隐私与可信度 .................................................................................19

1.2.7　接口 .................................................................................................20

1.2.8　算法 .................................................................................................23

1.3　机器学习与Netflix Prize .............................................................................24

1.4 MovieGEEKs网站 .......................................................................................25

1.4.1　设计与规范 .....................................................................................27

1.4.2　架构 .................................................................................................27

1.5　构建一个推荐系统 .....................................................................................29

小结 ......................................................................................................................31

第2章　用户行为以及如何收集用户行为数据 ..................................................32

2.1　在浏览网站时Netflix如何收集证据 ..........................................................33

2.1.1 Netflix 收集的证据 .........................................................................35

2.2　寻找有用的用户行为 .................................................................................37

2.2.1　捕获访客印象 .................................................................................38

2.2.2　可以从浏览者身上学到什么 .........................................................38

2.2.3　购买行为 .........................................................................................43

2.2.4　消费商品 .........................................................................................44

2.2.5　访客评分 .........................................................................................45

2.2.6　以（旧的）Netflix 方式了解你的用户 .........................................48

2.3　识别用户 .....................................................................................................49

2.4　从其他途径获取访客数据 .........................................................................50

2.5　收集器 .........................................................................................................50

2.5.1　构建项目文件 .................................................................................52

2.5.2　数据模型 .........................................................................................52

2.5.3　告密者（snitch）：客户端证据收集器 ..........................................53

2.5.4　将收集器集成到 MovieGEEKs 中 .................................................54

2.6　系统中的用户是谁以及如何对其进行建模 .............................................57

小结 ......................................................................................................................60

第3章　监控系统.........................................................................................61

3.1　为什么添加仪表盘是个好主意 .................................................................62

3.1.1　回答“我们做得怎么样？” ...........................................................62

3.2　执行分析 .....................................................................................................64

3.2.1　网站分析 .........................................................................................64

3.2.2　基本统计数据 .................................................................................64

3.2.3　转化 .................................................................................................65

3.2.4　分析转化路径 .................................................................................69

3.2.5　转化路径 .........................................................................................70

3.3　角色 .............................................................................................................73

3.4 MovieGEEKs仪表盘 ...................................................................................76

3.4.1　自动生成日志数据 .........................................................................76

3.4.2　分析仪表盘的规范和设计 .............................................................77

3.4.3　分析仪表盘示意图 .........................................................................77

3.4.4　架构 .................................................................................................78

小结 ......................................................................................................................81

第4章　评分及其计算方法............................................................................82

4.1　用户-商品喜好 ...........................................................................................83

4.1.1　什么是评分 .....................................................................................83

4.1.2　用户 - 商品矩阵 .............................................................................84

4.2　显式评分和隐式评分 .................................................................................86

4.2.1　如何选择可靠的推荐来源 .............................................................87

4.3　重温显式评分 .............................................................................................88

4.4　什么是隐式评分 .........................................................................................88

4.4.1　与人相关的推荐 .............................................................................90

4.4.2　关于计算评分的思考 .....................................................................90

4.5　计算隐式评分 .............................................................................................93

4.5.1　看看行为数据 .................................................................................94

4.5.2　一个有关机器学习的问题 .............................................................98

4.6　如何计算隐式评分 .....................................................................................99

4.6.1　添加时间因素 ...............................................................................102

4.7　低频商品更有价值 ...................................................................................105

小结 ...............................................................