The skills you have need maintenance and occasional updates. Doing an interesting data science project is what will keep you from getting rusty.
I believe in practice. Practice as an application of knowledge and ideas. One idea I don’t believe in is the one saying the road between theory and practice is a one-way highway. In other words, it tells us that practice is simply the application of theoretical principles. But practice is much more; it is also a birthplace of ideas that gives a push to new theories.
Reading articles or research papers, attending conferences or meet-ups to keep abreast with the latest technologies, and gaining better theoretical foundations is nothing to frown upon. I highly recommend that! But whatever the updates you get, your work as a data scientist will still boil down to several fundamental skills: data collection, analysis, and visualization.
And you need to use them! If you lack opportunities for that, you have to create them yourself. The most comprehensive way is to think of the data science project and do it from start to finish.
In such projects, you would use APIs to get the actual data. Through data cleaning and analysis, you would get insights, which you could present in some nice visualizations. Finally, you could post it on reddit, get the feedback and potentially take them into account to improve your project.
Ideally, the projects should also be fun, not only a drab way to polish your skills.
The project’s idea is to get the data about the top 10 most popular songs on Spotify through the Spotify API. The Spotify metadata can then be connected with the lyrics from Genius API or some other lyrics site.
The definition of the ‘word’ in the lyrics is up to you. For example, you can count the words in total or only unique words. Do you include singalong parts like ‘na na na’ or ‘la la la’?
By analyzing this data, you can show the historical development and predict future tastes. You could include some other parameters, such as the song length or the intro length. Especially the intro would be interesting to see because the studies, such as the one by The Ohio State University, show a dramatic decrease in the songs' intros length throughout the years.
You can take some ideas on the approach and the visualization from this project.
If you’re considering investing in properties across the country so you can rent them, it would be useful to analyze which factors influence the rental price and, therefore, the profitability and potential of your business.
The data can be acquired through some of the rental APIs. The factors that could come up as important in deciding where and when to invest could be the location, property size, date of build, amenities, rental price trends, etc.
The inspiration for the approach can be drawn from this nice Airbnb data science project.
Here you can use Facebook, Twitter, or reddit API to get the data you’ll work with. Based on the data you have, you can analyze the posts on social media and separate the fake news from the non-fake ones. Your approach could be more general, but you can also focus on something.