The ninth CDA Data Analyst Certification Exam ended successfully at the end of December 2018. Recently, we interviewed several excellent students who ranked among the best in this exam. Before that we interviewed the champions of Level 1 and Level 2 big data direction (click to view the previous interview). The top three in the direction of the model, see how they learn and prepare for the exam?

Level 2 Modeling · No. 1 Cai Xiaojing

Graduated from Beijing University of Technology in Information and Computing Science in 2017

Opportunities to apply for the CDA certification exam I do n’t have much data analysis in my current work. However, due to my major and data analysis related, I am also interested in this aspect, and I have the idea of switching to the future. Considering the gold content and professionalism of the CDA certification exam in the data analysis industry, I finally chose to apply for the exam.

How do I prepare for the test? In the preparation for the test, I think the most difficult thing is to balance work and study. After joining the work, I don't want to be able to quickly enter the learning state as in the student years. In this process, perseverance and willpower are particularly needed.
Preparing for the exam I started reviewing and preparing about three months in advance. First, I needed to make a plan and stipulate how much content I want to read every day. And it is important to adapt yourself to the learning environment in advance.
1. The syllabus is very important and the key knowledge syllabus is listed, and the syllabus is always the main one. You can print out the syllabus and learn while taking notes. After all, good memory is better than bad writing.
The following are my syllabus notes ~
2. Flexible use of time Because I am an office worker, I will use the scattered time between breaks to learn. After work every day, concentrate on time to overcome the more difficult knowledge points, spend about 3 to 4 hours per day to learn.
In the last two weeks, it is recommended to consolidate, understand, and fill in vacancies for knowledge points in the syllabus. Thoroughly understand each question and analyze each option.

Recommended courses

I signed up for the "CDA Level II Modeling Analyst" course, and I think learning with video is the most systematic. After learning a whole set, following the teacher's key notes.

Advice for test takers For students preparing for the test, I mainly have the following three suggestions.
1. It is important to lay a solid foundation. It will be relatively simple to understand the difficulties in the basic exam. If you only focus on overcoming the difficulties, you will lose more than gain.
Second, we must adopt the most suitable learning method. It is not possible to pass the surprise review before the exam. It is best to start the review in advance. After all, the plan can't keep up with the changes and adapt yourself to the state of learning in advance, otherwise it will be difficult to stick to it later.
3. Be confident. Don't scare yourself in terms of mentality, give yourself confidence, and on the basis of studying hard, the exam is not too difficult. In fact, I was preparing for the exam step by step according to the established plan, and I learned that I had a little surprise after taking the top prize (laughs). Therefore, we must give ourselves confidence and believe that we can.

The future career development plan hopes to apply the knowledge gained by data analysis to the current work, and may consider switching careers in the future.


At present, I am doing back-end maintenance work in Sichuan Telecom. I was originally responsible for the maintenance of the business platform. Now I am developing and developing big data. At present, I am trying to do correlation analysis of network equipment alarms.

The company that took the opportunity to apply for the CDA certification exam has previously organized some training on data mining analysis. The main purpose of applying for the CDA this time is to understand your learning situation and current level, and at the same time revisit the basic knowledge and sort out the knowledge structure.

How did I prepare for the exam? I participated in a week-long pre-test training in November, which basically started around CDA courseware.
One week before the exam, I started to make up the video course of teacher Li Yuxi. While following the video learning, I also refined the courseware and carefully mastered every knowledge point.
At the same time, there is a live broadcast course by Li Yuxi about 3 days before the exam. It is about the preparation before the exam and the students' questions and answers. It is still very important.

The recommended books and course exams are mainly based on the syllabus, and carefully study the courseware provided by the CDA official website and teacher Li Yuxi's course video. Don't miss any details.

System learning books:
"Data Mining Concepts and Technologies" (difficult to chew)
"Machine Learning" Zhou Zhihua's Watermelon Book (Slightly Good)

What are the knowledge and difficulties in the exam? The evaluation index for practical problems is not the traditional classification accuracy rate, but two results based on the profit index. I haven't touched this knowledge point before, combined with the video content of teacher Li Yuxi, anyway, research and understand, and complete the code.

Advice for test takers
1. First of all, we must systematically learn the process of KDD, in which data preprocessing and modeling are the top priority. The two complement each other and are a process of continuous iteration and improvement. It is necessary to understand the advantages, disadvantages and application scenarios of mainstream algorithms.
2. Be sure to fight in actual combat, not to talk on paper. No matter what tools are used, Python, R, etc., only through continuous operation through examples can we gain a deeper understanding of knowledge.
3. Not only to learn, but also to find a place to use. In the final analysis, data mining is just a tool, just like today's computers. How to use tools to solve practical problems is the most useful. Combining work and life, find application scenarios of data mining, so that the motivation for learning is more sufficient.
4. For the exam, be sure to watch the live broadcast before the exam. The teacher will talk about the details of the exam, which is not in the book. In addition, there is nothing to say about objective questions, which is the usual accumulation. Practical questions, please prepare the training model, mainly including missing value (discrete, numerical) filling, numerical discretization or normalization (combined model), data encoding, model selection (recommended strong forest, random forest, xgboost, SVM and other strong classifier )and many more. Although the actual operation has more than 2 hours, basically there is no time to analyze each feature attribute one by one. It is recommended to preprocess coarsely, score the model, select the appropriate model, and then try a more detailed preprocessing according to the selected model to see if the accuracy rate has improved.

I have just started my career development and planning data mining in the future, and there are many things to add. I hope to combine the actual problems at work, use the tool of data mining, and find a new direction to solve the problem.


The current job is currently graduated two years, the major is agronomy, because of interest in data, learn data-related knowledge after graduation. Previously working as a data analyst at an Internet finance company, he is currently in a state of resignation. This exam is also a charge in his own rest adjustment.

The opportunity to apply for the CDA certification exam hopes that I can supervise my study through the exam, and I also want to prove myself through the relevant certificate. Since there is no national unified exam, CDA is recognized by more and more professionals and companies, so I CDA is selected.

How do I prepare for the test? I am interested in machine learning. In the months before the test, I have self-study to consolidate the knowledge of machine learning algorithms. Twenty days before the test, I started to conduct a targeted review with reference to the outline, and used the online simulation data set for practical exercises one week before the test.

For recommended books and course algorithms, see Li Hang's "Statistical Learning Methods". In terms of tools, I prefer Python, and I recommend following the video course to learn and practice the code.

The advice to the test taker personally feels that the final evaluation standard of the model is very important. The evaluation indicators usually contacted are generally KS and Recall, but the evaluation standard used in this test is Profit. I haven't known this evaluation index before, so I spent a lot of time doing modeling exercises for this index before the exam. For different evaluation indicators, different data sets should adopt different modeling methods.

In the future career development plan, I am currently resting and adjusting. The previous work was mainly in data analysis. I hope to work in data mining in the future, so that I can apply more of the knowledge I have learned to practice and use machines. Learn knowledge to provide more feasible predictions, better evaluate customers, and reduce customer churn.