Projects are an essential part of learning data science. And, deciding on a topic for a project is a tough nut to crack. Here I have shared my experience & a three-step formula to find a personal project idea.
If you have decided to work on a project, think like a doctor. What does that mean?
Think about what are the things that can improve your day-to-day life with the help of data. You'll come across a lot of problems. And one such problem is your data science project.
Think of a problem you might be facing that can somehow be connected with data. Now ask the question, can you solve it with data?
If your answer is no. You can learn & revisit more topics in Data Science.
If your answer is yes, it means you have hit the right target and you are thinking in a suitable direction.
If still, you are not sure how you'll solve the problem i.e. what data-driven approach should be followed to solve the problem statement, you can follow the following 3 steps technique after figuring out the problem statement
1) Approach: You have to figure out what kind of method you should prefer to solve the problem. Does it require machine learning, mathematical programming, mathematical analysis or something more advanced
2) Data: Next step is to find the relevant data according to your problem. Is it available on the internet, or do you need to scrape it? Is the data structured or unstructured, and how do you clean and preprocess the data? These are some questions that you need to ask yourself to get quality data for your problem statement.
3) Result: Once you get the data and you have applied the chosen approach, it's time to present the solution to the general audience. Writing a detailed report of your findings is the best way to present your project. The report also helps the key persons to understand your project without going through each line of code.
This 3 step technique does wonders not only for personal projects but also for professional projects.
Let's understand by an example how to use this 3 step approach to a given problem statement.
Problem: How to increase subscribers of a YouTube channel?
(Disclaimer:
The youtube algorithm is far more advanced than the solution presented here. The solution is just to understand the strategy)
1. Approach: This problem may require machine learning techniques like Regression, Random forest, or advanced techniques like Neural Networks.
2. Data: There are many ways to get the data of a YouTube channel. We can use the YouTube API, or we can ask the owner of channel.
3. Result: The way we present the result is the most important thing. Saying something like, "make quality content for a youtube channel" as a result might not be an effective answer to the given problem statement.
The result should be specific, actionable, and personalised. Mentioning something like, "Posting twice a week, replying to all comments, uploading videos in the evening, etc are the key insights from the data & analysis, that have increased the engagement & retention in the past, so following these tips will help in gaining subscribers.'' will be very effective.
So, the overall summary for finding a suitable data science project is:
Look for problems - Can the problems be solved with the help of data? - Solve them using 3 step technique (Approach - Data - Result)
No comments:
Post a Comment