1. Briefly understanding of the user's portrait.
User portrait that users tag information, the enterprise through the collection, the analysis of user data, abstracted a virtual user, can be considered a virtual representation of real users.
2. Description of the main process building user portrait.
Basic data collection -> behavioral modeling -> Build portrait
3. personalized recommendation system performance can be judged by what criteria?
Customer Satisfaction coverage prediction accuracy cold start problem of excessive recommend personalized recommendations topical issues
4. Description of understanding of the CTR calculation formula.
It refers to the ratio of the number of ad impressions that are clicked open in the total number of impressions.
5. Factors affecting CTR of what?
(1) The impact of advertising itself, and the type of ad content, impact on traffic is very significant
(2) the context of environmental impact, the position of online advertising that appears extremely important.
Impact (3) Advertising viewer, different people have different preferences, which can lead to advertising "preference" different
6. Click ad estimated methods are there?
(1) direct estimation method
(2) Click to rate forecast model calculation method
7. briefly with the recommended location-based advertising forms.
(1) "Active", also called "push" type, refers to the advertising service provider based on the user's location, the initiative to send advertisements to customers until the user cancels the ad subscription or ad blocking so far.
(2) "passive", also known as "pull", refers to a user initiates a search by keywords, recommendation system returns results based on a recommendation keyword search, the user's current location information and other user features.
8. outlined the concept of Internet banking.
Internet financial means to pay for relying on cloud computing, social networking and search engines and other Internet tools, implementation of a new financial financing, payment and information services such as intermediary. Internet banking is secure, mobile and other networks on a technical level, the new model automatically after being familiar with the accepted user to adapt to new demands and new business generated.
9. DESCRIPTION application direction of the large financial data in the Internet.
(1) Fraud and Financial Analysis
(2) to build a more comprehensive credit evaluation system
(3) high-frequency trading and algorithmic trading
(4) public opinion analysis products and services
Big Data applications in finance 10. Description of machine learning.
(1) credit scoring algorithm
Performance Evaluation (2) Classification Model
Role in big data finance briefly 11. The machine learning.
12. Description of mainstream credit evaluation algorithm What?
（1）逻辑回归（Logistic Regression）算法
（2）支持向量机(SVM)
（3）决策树（Decision Tree）
（3）随机森林（Random forest）
（4）自适应提升(AdaBoost)
（5）梯度提升决策树(GBDT)
13.简述分类模型的评价体系。
（1）混淆矩阵（Confusion Matrix） -> 提升图（Lift），增益图（Gain），受试者工作特征曲线（ROC）。
（2）受试者工作特征曲线（ROC） -> 曲线下面积（AUC），洛伦兹曲线（KS） -> 基尼系数（GINI）。
（3）标准误差（MSE）独立出来。