5.1 Construction of the electricity supplier and user profiling technique processes
Construction of a portrait of the user, the data source comprising data collection, data preprocessing, behavior modeling, build a user portrait
Some tags can be sent directly to some label required by data mining analysis to!
5.2 Data Analysis source 
user data is divided into two types: dynamic information data, the static information data
Static information Source:
Ø users fill out personal information, or result through a certain algorithm, calculated data
Ø If there is uncertainty, can be modeled to determine, such as the user's gender did not fill out the registration, can be modeled to determine what the user based on the user's behavior gender, or its probability
Dynamic Information Source:
Ø user behavior data generated: registration, visit, click, purchase, receipt, evaluation, collection and so on.
Ø user behavior data more important: Tour merchandise, collection of goods, Add to Cart, Watch item
These behaviors can be calculated based on the characteristics: user registration time, the first time a single, tidal mother family, tangled goods, the largest consumer, the number of orders, the number of returns, ruining the index, brand preference and so on.
5.3 Analysis of target
User portrait goal is to analyze user behavior, and ultimately to tag for each user, as well as the right to re-label.
E.g., red wine 0.8, 0.6 lining.
Tags: the performance of the content, the content users are interested in, preferences, needs, and so on.
Weight: the performance of the index, the user's interests, preferences index may also show the user's needs, it can be simply understood as the credibility of probability.
5.4 Modeling 5.4.1 User Basic user portrait attribute table
According to the user to fill in the Properties tab and extrapolated label. To understand the basic situation of the population and the statistical properties of the user according to different attributes dimension.
Role: Marketing by demographic attributes, such as post-marketing 80, preferential Taurus, birthday marketing.
The main sources of data: user tables, user survey, pregnant women, model table, vest model table.
Users Table: record the user basic attribute properties.
User survey: complement other user's basic information.
Users fill in the basic information: user ID, user name, password, gender, phone number, email, age, household provinces, ××× number, registration date, delivery address, etc.
User information filled calculate the index obtained by:
Before the birthday, constellations, city level, several mobile phones, mobile phone operators, mail carrier
Users get the questionnaire: education, income, occupation, marital status, whether you have children, whether a car has room, the use of mobile phone brands.
According algorithm:
身高、体重、性别模型、孩子性别概率、潜在汽车用户概率、是否孕妇、孩子年龄概率、手机品牌、更换手机频率、是否有小孩，是否有车，使用手机档次，疑似马甲标准、疑似马甲账号数、用户忠诚度、用户购物类型。
模型算法---性别模型
Ø 用户自己也填写了性别，但仍然要用算法算一次性别
Ø 性别验证方法
随机抽样几千条数据让客户打电话确认。
与用户自己填的性别做对比，确认百分比。
模型算法---用户汽车模型
模型算法---用户忠诚度模型
Ø 忠诚度越高的用户越多，对网站的发展越有利
模型算法---用户身高尺码模型
模型算法---用户马甲标志模型
Ø 马甲是指一个用户注册多个账号
Ø 多次访问地址相同的用户账号是同一个人所有
Ø 同一台手机登陆多次的用户是同一个人所有
Ø 收货手机号相同的账号同一个人所有
模型算法---手机相关标签模型
Ø 对于手机营销参考意义比较大
Ø 使用手机品牌: 最常用手机直接得到
Ø 使用手机品牌档次：根据档次维表
Ø 使用多少种不同的手机：手机登陆情况
Ø 更换手机频率（月份):按时间段看手机登陆情况
5.4.2 客户消费订单表
根据客户消费的情况提取的客户标签，用于了解用户的消费总体情况，
最终的目的根据用户消费习惯与消费能力做营销。
主要数据来源：订单表、退货表、用户表、购物车表
订单表可以得到相关标签：
退货表可以得到相关标签：
购物车表可以得到相关标签：
订单表和用户表可以得到相关标签：
5.4.3 客户购买类目表
根据客户购买类目的情况提取客户标签，用于了解类目的购买人群情况和针对某一类目的营销等。
主要数据来源：订单表、购物车表、类目维表
类目维表可以得到相关标签：
电商的三级类目：
京东商城：
淘宝：
Orders tables and dimension tables can be related to the category labels:
Cart tables and dimension tables can be related to the category labels: 
5.4.4 users to access information table
Extract relevant customer label according to the customer visit.
To understand the user's overall access and easy to do marketing based on customer habits tour
Main Source: clickstream behavior log table (PC / APP end)
Clickstream behavior log table can get related tags: