The 9th CDA certification exam ended successfully at the end of December last year. According to international practice, we will interview after each exam The top candidates in the exam share their test preparation experience and journey. Then the interested friends may have noticed that in the previous No. 1 interview, only Level 3 is vacant. Why? We all know that CDA certification is divided into CDA Level 1, Level 2 and Level 3, among which Level 3 data scientists have the highest gold content and the most difficult exam. No one has been able to pass it. However, in the ninth CDA certification exam, the only Level 3 passer was ushered in. He was also the first data scientist to pass the CDA certification!

So what kind of "Great God" can pass the Level 3 exam? Today let us walk into Zeng Jin, the first data scientist in CDA Level 3, to see how he prepares and learns step by step, and finally passes the highest certification of the CDA certification exam.


Senior data product manager of Qunar.com, master of finance from Central University of Finance and Economics. At present, he is mainly responsible for the construction and application of the whereabouts ticket service platform, the user portrait of the user product department and the BI system. He has many years of data analysis and practical experience.

Q1: Please tell me about your educational background and my current job. I am not from the data analysis class, or even a liberal arts student. Bachelor and Master graduated from Beijing Technology and Business University and Central University of Finance and Economics, respectively. The majors studied are finance in economics. It is a liberal arts-oriented major. It seems that it has nothing to do with big data. Later, after participating in the work, he has worked as a researcher and business analyst in consulting companies and game companies. The experience of data analysis and data mining is accumulated in the study and practice of later work.
Currently, I work as a senior data product manager in the Qunar ticket business department of Qunar. I manage a data product team of more than ten people. Mainly responsible for the user portrait and modeling, data analysis, BI system construction and other work of the Qunar network ticket. At work, the data analysis skills we use are mainly in three aspects. One is data mining modeling. Our data products mainly use R language and python for data analysis and modeling. The second is business analysis skills. The third is data warehouse and data acquisition skills, SQL, Hive, etc.

2. What is your chance to apply for the CDA certification exam?
At present, there is no national unified examination certification for big data, and CDA has a relatively perfect design for the data analyst's skill system. In particular, CDA Level 3 requires comprehensive, and you can improve your knowledge system and skill tree through the process of exam review.

3. Please talk about your mental journey of the exam, such as how to get from Level 1 / Level 2 to Level 3 step by step.

I took the CDA Level 2 modeling analyst exam two years ago, and I'm no stranger to CDA. At that time, by preparing for the exam, I helped myself to find out and fill in the vacancies to improve the knowledge board, and benefited a lot.
Therefore, we have been paying attention to the registration information of CDA Level 3. I signed up as soon as the CDA Level 3 exam opened this year.
The requirements of a data scientist are comprehensive. During the review process, I deeply found my deficiencies in theoretical knowledge and also repaired my skill tree through review. Moreover, the amount of knowledge required for the data scientist exams also improves his ability to reasonably arrange time and fight stress.
The second stage of the CDA Level 3 exam is part of the case exam. I am honored that teacher Li Yuxi is my answering examiner. Teacher Li Yuxi was kind and careful, gave me a careful and patient evaluation of my case, and pointed out the deficiencies in my modeling process. I hope that I will pay more attention to the interpretation and application of the model, which will benefit me a lot.

4. You are preparing for the exam, how do you balance work and study arrangements?
Since the work is usually busy, the time I can use is only 1.5 hours a day from 10:30 to 12:00 in the evening and the time on the weekend. Therefore, I have done some homework in priority scheduling and learning in stages:
(1) Prioritization I first arranged the priorities and review strategies for the content in the outline based on my own basis (as shown below).
For example, I am relatively familiar with most of the content of machine learning and deep learning, and this part of the score is close to 50? It is also the key content. Corresponding to the first quadrant above, it needs to be fully read through and leaks filled. Through learning and combing, I have a more systematic understanding of SNA, reinforcement learning and category imbalance.
Computer science and technology (score 15? And big data architecture (15? Is something I am not familiar with (because I am a data product sequence rather than a data development sequence), corresponding to the second quadrant above, which requires key research It took me more than 40 hours and energy to prepare for this part of the exam.
There are also applications such as deep learning in face recognition and object monitoring. The score is only 1? I am not very familiar with it. If these contents are to be studied in depth, it takes a lot of time. It belongs to the third quadrant, so I understand that can. Do not drill the horns.

(2) Learning in stages
CDA Level 3 has more content, it is best to start the review at least three months in advance. I personally started the formal review in September.
The first stage: The first round of review (2018.9-2018.11) reads and digests the knowledge in the book, and each part of the outline forms a brain map, which is convenient for you to master systematically.
The second stage: Thematic breakthrough (from November 11, 2018 to mid-2018.12) Focus on the weak subject. The third stage: Sprint review (from mid-December 2018 to the front) is mainly based on memory.

5. In the process of reviewing CDA Level 3, what recommended books and courses do you have to closely follow the exam outline. You can combine the outline with the online resources to read the classic bibliography, and you can also pay attention to some CDA official live broadcasts and videos (teacher Li Yuxi had a live broadcast course in the direction of "machine learning" before the exam, which is consistent with the exam content).
"Deep Learning" by the three great cattle of Goodfellow, Bengio and Courville, must be intensively read.
Teacher Zhou Zhihua's "Machine Learning" is easy to understand and has a complete system. You can learn according to the content required in the outline (the machine learning part of CDA Level 3 and the machine learning part of the CDA Level 2 modeling analyst are quite different, so I am reviewing Only look at the parts required by the CDA Level 3 outline)
In addition, "R High Performance Programming" and "Social Network Analysis Methods and Practices" have greatly inspired my daily work.

6. Advice for test takers
CDA Level 3 has high requirements for the solidity and comprehensiveness of the candidates' theoretical foundation. The content covers a wide range. There are dozens of reference books, which are more than one foot thick ... So there are three suggestions to share with you:
(1) You must thoroughly understand the requirements of the syllabus. The syllabus should not be required. You must give up. Time is very limited and precious! You can allocate your time and energy according to the percentage of points given in the syllabus.
(2) Have a study plan and control the learning rhythm. For example, if you plan to read a certain chapter of a book today, you must complete it; find a good rhythm, such as the first round of review, the second round of review and the sprint review (more like the college entrance examination).
(3) Pay attention to the learning method. During the through reading stage, I will draw the content of the book into a brain map to help me better remember and understand the correlation in the knowledge points; Recording, put twice the speed listening in the fragment time, deepen the impression.
The above is my preparation process, and I hope that candidates will find the most suitable method for them in the exam and study. It may be difficult at first, but as long as you keep your mind straight and stick to it, I believe you will definitely achieve the desired results!

ps. Last year I insisted that I lost 40 pounds in running fitness. This may be one of the most accomplished things I did last year (laughs).