Overview
- Winning data science contests can be a complex procedure, but it can reach the top 3 if you have a framework to follow.
- Hear from a data science hackathon expert and how he went from scratch to winning data science contests
Introduction
There is no alternative to learning through experience. Especially in the data science industry!!
I recently won first prize in Zindi's Zimnat Insurance Referral Challenge, an achievement that ranks first among my data science competition results of all time.
In pure numbers, this was not my first top result, but only one of the more than 30 results between 3 first I've had on my own data science competition journey. During this period of starting from scratch and ranking in the higher echelons of the machine learning hackathon leaderboards, I have realized the relevance of learning through experience and I cannot stress enough how important the above quote is.
Winning a data science contest is a hassle-free journey. You are competing against the top data science minds everywhere!, you're working on a data science hurdle that hasn't been solved before and you're doing all of this with a strict deadline to boot!
But I can assure you that getting the 3 first places in the leaderboard is absolutely achievable, if you know what you are doing.
It is not intended to be a technical publication.. It's about my journey to data science competitions. Y, of course, how and why should you start right now. As a beginner, you certainly don't want to miss reading this. Technical posts will appear soon for more advanced readers, So stay tuned!
As I mention, there is no better way to learn data science than by practicing it. I encourage you to join us this extended weekend for a guided community hackathon where data science proficiency experts will take you through the entire hackathon procedure LIVE:
And you can always visit the Plataforma DataHack to practice your data science skills or participate in hackathons.
My journey into the data science competition: from zero to expert
One of our professors introduced me to data science at the beginning of the third semester at university.. I was using Machine Learning to discover Earth-like planets and the chance for extraterrestrial life.
Curiosity followed me and led me to dive into Andrew Ng's famous course on machine learning. I was introduced to various applications of machine learning, like the stock market price forecast and autonomous cars, to name a few.
Searching more in the Google search engine about potential possibilities in this field, I discovered platforms like Kaggle and DataPeaker. It added more fuel to my growing interest in data science. Constantly competing and improving against time and a leaderboard was the next challenge; Yes, I'm talking about data science hackathons!
Most of the beginners I've interacted with feel like they need to know the ins and outs of machine learning first.. Only then can you start competing in Data Science.
That's a big mistake.
“To take part in data science competitions, you just need the drive to constantly learn and improve. You will get a good ranking “.
My first competition: Kaggle's Microsoft Malware Prediction Challenge
Let me quickly talk about my first serious competition at Kaggle.: Microsoft's Malware Prediction Competition. This came months after failing a range of data science contests. But the experience gained in all the competitions up to that point had helped.
In solo 2 weeks and with some presentations in hand, I jumped to the top 20 of the public classification.
As time went, I teamed up with a Singaporean student, a master from Kaggle and two London industry leaders, New York and Pune. Working together in different time zones was a challenge in itself, but we managed to discuss and implement strategies and models day and night in Slack.
And in conclusion, with me leading the team, we ended up in the post 25 in private classification. This was pretty close to our public ranking of ranking of 21. It was a very good result, whereas only 10 teams between 100 first in the public ranking were able to maintain their position in the private ranking.
Fast forward to the current day, I have finished between 3 first in more than 30 data science hackathons on various platforms. This includes the number one position on almost every major platform that I have participated in. (and if, Top two in DataPeaker's JantaHack series).
This is a brief summary of my journey to conquer data science competencies from scratch. Then, let's understand how can you, as a beginner, start participating in data science contests.
How do I start data science competitions if I am a beginner?
Here's some advice I wish someone had given me when I started competing in data science hackathons: enter any competition you feel comfortable with. The most important thing is that you start.
Vidhya's analytics JanataHack is a series of competitions for beginners held every week. In the end, many winners are also kind enough to publish their solutions.
Anyone just starting out should make sure they go through winning solutions for past data science contests. When you come across a new idea or concept, search it on google search and take the time to understand it. If you are unable to transfer your learning from one competency to another, you haven't used your time properly.
Transferring learning is very important, from deep learning to learning.
How do I approach data science contests?
Here, I have written some key tips that you should pay attention to when starting a new data science competition..
- I regularly start with a simple baseline model. Just take a look at the data, then create a model without any data cleaning or feature engineering
- Then, the goal becomes understand the problem and the data to create a good validation set. A good validation set is a must. Only then can you trust your local results. Opposite case, get ready for a private leaderboard reorganization.
- Feature engineering is the next key step. Good characteristics always differentiate between a winner and a top 100
- As the end of the competition draws near, i usually try to build a range of models like gradient increase models, neural networks, etc. Then follows the stacking or combination of these results. Assembly gives you the edge to win a competition. Therefore, it is a tool that you will always want to have on hand.
- One thing that a lot of people don't talk about is the relevance of a code base. Time is a very important factor in any data science competition. You shouldn't waste time writing the same snippets from scratch over and over in multiple competitions. Instead, focus your precious time on doing something new and better
What are the benefits of participating in data science contests?
That is a valid question!! Data science contests require a significant amount of your time, then, they're worth it? Let me share some benefits of my experience in this section.
1. Compete and learn
Learn a lot during data science competitions, from troubleshooting to model building. If you intend to learn something new, contests are the best way to do it. Soon, will study and experiment a lot, and you will constantly find yourself looking for better alternatives to improve your model.
2. Networking
To this day, I have partnered with more than 25 different people from india, Singapore, EE. UU., England, France and Africa in different data science competitions. These people range from students to industry leaders.
Honestly, networking is one of the biggest benefits of participating in these hackathons. Meeting and interacting with like-minded people is undoubtedly a great asset to your future career..
I got my current job at DataPeaker thanks to networking!!
3. Profiling / Resume creation
Imagine a scenario where you are hiring a data scientist and you have shortlisted two great candidates. Both people have similar backgrounds in data science. The first person has completed some projects in data science, while the second person has completed similar projects, as well as “Achieved rank” X “in a data science competition competing against hundreds of people”.
Then, Which one would you like to give more preference to? As a hiring manager, most people would prefer the second option.
This is not to undermine the relevance of a good project., but a good ranking in a data science competition definitely gives you an edge over your competition. Nowadays, many companies prefer candidates with experience in data science competencies. As an aspiring Data Science, It's time for you to start also!
4. Get rewards and win exciting prizes
Finally, but not less important, seasoned data science competitors have a lot to gain and gain. Only during this confinement, I made enough money to buy a car. Platforms like Kaggle have a lot for you if you think you have the ability to solve the world's most interesting data science problems.. What are you still waiting for?
HackLive – Guided community hackathon!
What if there was a live session that could encourage and help beginners to participate in data science hackathons and improve their ranking?? Wouldn't that be great?
Since its inception, DataPeaker has been trying to decode the problems facing the data science community and present a viable solution to them.. And the inability to start participating in Data Science Hackathons has been frequent.. Then, as a step to address this issue, let me introduce: HackLive 2 – Guided community hackathon!
DataPeaker data scientists will combine all of their industry experience and knowledge to help the community respond 3 questions:
- Is it important to highlight if I have a minimal chance of winning?
- How do I start?
- How can I improve my rank in the future?
Then, What are you waiting for? Go and register in the link below:
Final notes
I hope I have given you enough motivation to start your own journey towards data science competitions.. More technical posts on upcoming data science competitions will be published soon. I am excited to share them with you! Until then, you can start with some of my hackathon solutions on Github here.
Are you a beginner looking for a place to start your data science journey?? Here is a complete course, full of data science knowledge and learning, selected just for you to learn data science:
Have you participated in data science hackathons before? How was your experience? Share your thoughts with us in the comment section below and we'll pick the best ones!!
