Introduction

** 12/13 Added each time and study time! ** ** ** 12/18 The part up to becoming Kaggle Master has been revised! ** ** ** 12/20 Added some parts! ** **

Purpose of the article

In this article, I'll look back at what I've done since I started studying programming until I became a Kaggle Master. I'll also share some of the things that made me happy to be a Kaggle Master, so I hope it's an opportunity to motivate you to work on Kaggle.

Career introduction

When I first started Kaggle, I was wondering what kind of background the person who became Kaggle Master had and what he had studied, so I will briefly introduce it!

I'm 21 years old, a third year undergraduate student, but I think I'm a relatively hobbyist. When I was a high school student, I belonged to the photography club and liked cameras, and I got Kendama Test Level 1 and played cards called Cardistry. I was playing with it. When I think about it, I think I'm relatively good at learning something by myself. (I also like photos that show my hobbies well and fountain pens)

Before entering university, I was interested in probability statistics and aimed to become an actuary who was active in the insurance industry. Therefore, when I was in my first year of undergraduate school, I mainly studied economics and actuary mathematics, but I feel that this experience is also useful when analyzing data.

In this way, until the summer vacation of the second year of undergraduate school, not only machine learning but also programming is not done outside of class (I was not good at it), but I wish I could handle data efficiently in order to play an active role as an actuary. With that in mind, I decided to use my summer vacation to study Python. Since then, I have been interested in machine learning and have continued to study until today.

From starting to study Python to participating in Kaggle competitions

Work on PyQ (2018/8 ~ mid-September 9 4h / Day)

First of all, I wanted to study basic grammar comfortably, so I searched for a site where I could study online. I met PyQ there. I'm really grateful that I learned the basic grammar through this site, but also realized what machine learning is and that it's an area of my interest. Perhaps I wouldn't have tried Kaggle if I hadn't studied on this site.

I'm tired of trying Udemy and Coursera (late September 2018)

Knowing the benefits of online learning at PyQ, I decided to work on Udemy and Coursera. I was planning to deepen my understanding on these sites, but as I was working on it, I quickly became more motivated to participate in the competition and gain experience, so I stopped halfway through. (Currently still in the middle ...)

Participation in the first competition (SIGNATE) (late 2018/9 ~ 2018/11 20h / Week)

Participating there was Student Cup 2018, which was just held at SIGNATE instead of Kaggle. In general, SIGNATE does not publish the Kernel, so you have to write the code from scratch on your own. However, in this competition, there was a person who published the baseline, and it was very helpful because I was able to work while studying what I was doing with that code. Personally, in such cases, there are no language barriers, so it's a good idea to start the competition with SIGNATE.

Also, looking back, I think it is one of the shortcuts to participate in the competition before studying until you are satisfied. Once you participate in the competition, you can acquire unknown knowledge through practice because you want to improve your score somehow. I feel that this is a very important point, and more recently, I am acting with a stance of trying to participate in this competition because I want to acquire this kind of knowledge. As a result, I was able to get into the top ranks, and it was a big harvest that I found it fun to participate in the competition.

Purchase a desktop PC (late October 2018)

I'm completely addicted to machine learning, so I've come to want to prepare a GPU and start learning. So I wanted a desktop PC, but I was really worried about how much to invest. I had a belief that ** I would buy the best I could buy at that time **, so I made it about 170,000 yen. I remember thinking at the time that this was a pretty big expense and I refused to retire: sweat_smile:

However, looking back now, I think it was a very good decision because I was able to earn many times as much money as the investment amount through internships and prize money. We will continue to do our best to invest as much as possible.

Participated in the after-event of the competition (mid-December 2018)

Fortunately, the Student Cup had an after-event after the competition. By participating in this event, it became clear what the difference was with the top players, and above all, it was great that I was able to give advice by talking directly. Also, at that time, there were no people around me studying machine learning, so I was able to increase my motivation by talking technically.

Read all the technical books (2018/10 ~ 2018/12, dozens of books)

Fortunately, there were so many books about machine learning in the university library, so I borrowed a lot and read them. At that time, it was like copying the source code that was posted with almost no knowledge, but I was able to gain the knowledge that there are such libraries and frameworks for such processing, and Kaggle I think there is no resistance when reading the Kernel. At that time, I think that there were few technical books such as this if you do Kaggle, but now ["Technology of data analysis that wins with Kaggle"](https://www.amazon.co.jp/dp/B07YTDBC3Z / ref = dp-kindle-redirect? _ Encoding = UTF8 & btkr = 1) is overwhelmingly recommended. Also, ["Theory and Practice by Python Machine Learning Programming Expert Data Scientist"](https://www.amazon.co.jp/%EF%BC%BB%E7%AC%AC2%E7%89%88%EF % BC% BDPython% E6% A9% 9F% E6% A2% B0% E5% AD% A6% E7% BF% 92% E3% 83% 97% E3% 83% AD% E3% 82% B0% E3% 83 % A9% E3% 83% 9F% E3% 83% B3% E3% 82% B0-% E9% 81% 94% E4% BA% BA% E3% 83% 87% E3% 83% BC% E3% 82% BF% E3% 82% B5% E3% 82% A4% E3% 82% A8% E3% 83% B3% E3% 83% 86% E3% 82% A3% E3% 82% B9% E3% 83% 88% E3% 81% AB% E3% 82% 88% E3% 82% 8B% E7% 90% 86% E8% AB% 96% E3% 81% A8% E5% AE% 9F% E8% B7% B5-impress- top-gear% E3% 82% B7% E3% 83% AA% E3% 83% BC% E3% 82% BA-ebook / dp / B07BF5QZ41 / ref = sr_1_4? __mk_ja_JP =% E3% 82% AB% E3% 82 % BF% E3% 82% AB% E3% 83% 8A & keywords =% E6% A9% 9F% E6% A2% B0% E5% AD% A6% E7% BF% 92 & qid = 1575727224 & s = digital-text & sr = 1-4) It was great because I could learn various methods.

From participating in Kaggle for the first time to taking solo silver (from January 2019 to the end of February 2018 25h / Week)

Of course, I also made my Kaggle debut at Titanic, but I saw Kernel a little and ended up moving it lol After all, I think it makes me want to participate in a competition where the leaderboard fluctuates in real time. Therefore, I decided to participate in the Elo Merchant Category Recommendation that was just held during the spring break.

This competition is about credit card information, and since it was a financial system that I was originally interested in, I was even more motivated. If you've tried Titanic for the time being, but don't know what to do next, you can enjoy ** participating in a competition in your area of interest **.

Since this competition was a table data competition, I was able to make use of what I learned at the Student Cup. ** It is a standard practice to use the knowledge of the competitions I participated in before to challenge new competitions **, so I feel that it is very important to participate in as many competitions as time allows in order to improve quickly. I will.

Also, in the article I read at that time, there was an advice that ** if you follow the discussion firmly during the competition and implement it, you will get a silver medal **. When I put it into practice in this competition, I was really happy to get the silver medal.

Until becoming Kaggle Master (2018/4 ~ late June 2018 25h / Week)

At the end of spring break, I became interested in natural language processing. The Jigsaw Unintended Bias in Toxicity Classification, which was held just at that time, gives you access to the latest technology. I thought it was perfect and decided to participate with all my might.

At this stage, I had no experience of participating in competitions with Keras or PyTorch, and there were many things I didn't know, so I decided to deepen my understanding by posting rather than just reading ** Discussion. ** ** In this way, if there are other people who are interested in the content, upvote will be attached and you can also get a medal, which is two birds with one stone. In Kaggle, medals often refer to Competition medals, but I think it's pretty amazing to be able to win the Master title with Kernel and Discussion medals.

Before I posted, I was worried about speaking in English, but when I got opinions from other users and came up with a solution, I felt that Kaggle was a really good community. Since then, I have come to want to answer questions and contribute as much as possible.

In this way, by deepening my understanding and implementing some original ideas, I was able to move up to around 20th place on the leaderboard, and I was able to form a team with people who were close to each other. Working as a team is a lot to learn, and I think it's full of fun that you can't enjoy solo. (It's a good memory to choose the final submit while consulting all night on the final day lol)

As a result, I was able to win the gold medal I longed for and had a very good experience. I feel that the key was to try something that I had never actively done **.

Specifically, I tried the following. ** 1. Created a topic in Discussion so that you can get the information you want efficiently. ** ** ** 2. Formed a team to acquire ideas that I couldn't think of on my own and at the same time motivated me. ** ** ** 3. I looked up related papers on arXiv and implemented the idea. (I felt it was important to work as much as I could.) ** ** 4. I read all the comments of Kernel that are attracting attention as well as Discussion. ** ** ** 5. I learned how to use it to turn learning on GCP, and later wrote Article. ** **

Behind my efforts, I met a friend who can talk about Kaggle and participated in a reading session in another laboratory, so I am very grateful for such an encounter.

After that, I won a silver medal and was able to become a Kaggle Master.

afterwards

Even if I went to the hospital, I wanted to know my evaluation at the moment, so I had an interview with several companies. At that time, I got a good response when I told them that I was a Kaggle Master, and sometimes I was treated like a graduate. I haven't belonged to the laboratory yet and have no research achievements, but I'm glad I've been working on it because I can introduce Kaggle as an achievement.

Also, when I started studying programming, I was given a freelance job that I hadn't expected, and I was able to gain valuable experience.

So, from now on, I would like to make further efforts to meet that expectation.

bonus

** For those who want to start Kaggle from now on **

You can get a lot from Twitter

Personally, creating an account for machine learning on Twitter is very useful in terms of collecting information. Also, it is highly recommended because you can increase your motivation by seeing the activities of strong people. Also, when I participated in Kaggle Days Tokyo, many people knew it on Twitter, so I enjoyed it very much.

in conclusion

It was my first time to look back on my efforts and write an article, and it was a mess. .. .. I would appreciate it if there are any points that were helpful. If you would like to know more about this part, please feel free to comment.

** Honestly, I really want a Macbook Pro: no_mouth: (I'm using the 2015 model now, so I'll write an article about building an environment for Kaggle) **

[PYTHON] Until and after becoming [Kaggle Master]