A Guide To Getting Data Science Project Ideas
How to come up with self-study, portfolio or business ideas. From someone with too many.
When writing about learning or breaking into data science, I always advise building projects.
It is the best way to learn as well as showcase your skills.
But I often get messages from readers asking, “How exactly do I come up with ideas for my projects?”
Any seasoned entrepreneur or engineer will tell you they have too many ideas. But it’s not always easy when you’re starting out.
So here’s a few ways I’ve personally come up with ideas.
Attend networking events and talk to people
Most people are surprisingly willing to share their own ideas. You just have to ask.
My default question at networking events is, “What are you working on or trying to solve?”
Last week at a virtual event, every single non-technical person I talked to shared a use-case for ML that they wanted to build.
Now don’t steal anyone’s idea. But if you’re already dedicating hours to learn data science, consider helping someone for free. You’ll get experience to put on your resume and a connection that may be useful in your career.
Successful people are happy to share ideas. They understand there are an infinite number of problems to solve in the world, and sharing isn’t a zero-sum game.
Use your hobbies and interests to generate ideas
Many great ideas have come from merging expertise across different domains.
For example, Geoffrey Hinton, the inventor of neural networks, had a background in psychology from which he drew many early ideas about artificial intelligence.
How can you apply this to your own interests?
Personally, I love my dog, badminton, and cooking. I’m also aware of the general topics under the machine learning umbrella. So I’ll try to match a type of ML with each of my hobbies to generate an idea.
- My dog — Categorize audio recordings of my dog’s different barks, ruffs and growls with machine learning.
- Badminton —Detect if a video of someone swinging a badminton racket has proper form, using machine learning.
- Cooking — Classify images of food, by country.
These could all be very interesting projects, if you dug deep into them.
So ask yourself, what are you interested in? Could data science help you do it better, or extract interesting incites?
Solve problems in your day job
Your current job may not be in data science. But that doesn’t mean there aren’t interesting data science problems to solve.
Every company has manual operational tasks begging to be automated. If you don’t have them yourself, your colleagues in marketing or customer service might. Can you help them?
Consider if automation, decision trees, or data visualization could help someone in your organization.
If this is outside your normal scope, you might have to work on it during your own time. But that’s a small price to pay if it adds value and gives you experience.
Back when I managed business intelligence for an e-commerce company, I wanted to break into software engineering. So I started writing code on weekends to scrape competitor websites selling similar products, and auto generated reports on our overpriced products. Then I sent the reports to our buying department so they could lower prices — This project helped me land my next job.
Go deep into your current job and you’re almost guaranteed to find a project that data science can be applied to.
Get familiar with the data science toolkit
Even if you don’t know how every model works, it’s valuable to know the general topics under the ML and data science umbrellas.
This gives you the ability to fit these models onto the world around you.
For example, I know that NLP encompasses “text classification”, “information retrieval” and “question and answer systems”.
So when I have a dataset in mind (ie: Reddit threads), it’s easy to think of potential applications and generate preliminary ideas.
Once you have the high-level toolkit, coming up with ideas becomes easier across the board.
Solve your own data science problems
What problems do you have in your search for a data science job? Could machine learning assist you?
Maybe you could scrape job boards, classify whether a job is data science related, and perform analytics on the job requirements.
That would be an awesome project!
You could also add competitive analytics showing hiring differences between companies, and show it to the company you want to work for.
As someone who hires engineers, I’d be fascinated to see the results of a project like this in someone’s portfolio.
Look at the world through data scientist glasses
Ask yourself what can be analyzed, tested, or automated as you walk around in your daily life.
Watering houseplants: could you analyze soil moisture to optimize plant growth?
Shopping: could the department store detect theft with machine learning?
Cooking: could a photo of the inside of your fridge detect what ingredients need to be replenished?
Then take the smallest component of the project, and actually try to build it.
There are an unlimited number of ideas to stumble across. You just need the right mindset to see them.
Conclusion
Coming up with ideas when you’re starting out is hard. I know because I used to be there.
But understand — all great ideas come from real experiences. There are no ideas in a vacuum.
That’s why it’s important to put down your laptop, get outside and talk to people.
Seasoned entrepreneurs have too many ideas because they’re already working on lots of projects, and cross-pollinating ideas between different domains.
Eventually, you’ll also get to the point where you have too many ideas. When you get there, share some!
