Background of the project
In recent years, along with fast Internet finance; the state has adopted the relevant documents, called for increased Internet trading risk prevention and control efforts; encourage large data analysis, establish and improve the trading risk detection model user behavior modeling and other means. But now there is a large risk control data, poor timeliness, accuracy not high. 综合用户分析平台包含 综合数据分析|登陆风险|注册风险|交易风险|活动风险分析等模块. The following is a relationship between the various subsystems.
- Service system : refers generally to a background or Web terminal APP + (target user service), the service module is the carrier, a large source of data calculated data.
- Wind control system (flow calculation): Support (service goal is to back-office services) for the business system, according to business data is more data collection system or in some buried Buried page to get user environment or user operating behavior.
- Punishment system : The system calculated based on risk assessment report, to control or punish the report with specific punishments business generated by the incident. For example: increase the verification code, prevent users log in, register or order.
- Comprehensive analysis :? Auxiliary air control system, based on the performance of sub-control system to determine the risk control strategy parameter settings or if there are problems. For example: assessment can effectively help operators and analysts factors trigger rate is very low or very high, etc., the system is completed discover new strategies and marketing improvements.
Platform Architecture
System Model
Real-time business
Sign up risk
检测出一些不正常的注册行为，并产生注册评估报告。
- In a short time with a device with a network of more than N number of registered data
- If the registration time is less than m min, and the behavior produces multiple registration, registration has identified abnormalities.
- The same device or IP, SMS to call several times within the time specified | e-mail service more than N times, the risk register
Login risk
检查出非用户登录，或者账号被盗窃等潜在风险挖掘。
- Remote login identified risk (not users often landed)
- Displacement speed than normal login, identified risk
- Replacement of equipment, there is a risk identified
- Login time - habits, there have been exceptions, there is a risk
- The cumulative number of logins per day exceeds the upper limit, there is a risk identified
- Enter the password input error or larger differences, the risks identified
- If the input characteristics of the user (the length of each control input required ms) change
Trading Risk
辅助系统检测用户支付环境，地区、网络、次数
- Users trading habits, abnormal periods identified as having abnormal
- Detecting a user's network environment, if not often used WI-FI
- Generate multiple transactions within a short time, there is a risk
- The payment amount is too large, there is a risk identified
- Unfamiliar area payments, there is a risk identified
- Trading equipment changes
Activity Risk
活动风险涉及活动时间、奖励兑换/兑换率
- Time ahead of the end, have identified risk.
- Events within a short time the exchange rate is too high.
- Repeatedly identified as having an account convertibility risk.
· Offline Business
辅助流计算或者企业运营部更好经营业务。
-
User analysis platform components (gender, equipment, age, region)
-
Application usage (time category | channel)
-
Click the channel or category list, stay long when the page, bounce rate
-
Wind control system to evaluate various types of application performance factor for each application in
