How To Learn DS PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

How to learn Data Science

1. Get a hold over the language to be used for implementation – practice the basics till you are
comfortable.
2. Practice and become comfortable with data cleaning and processing – this is essential. You
need to know about your data well before analyzing it.
3. Understand and implement one algorithm at a time. You may not understand it completely in
the beginning. Give it time. Do not get stuck at one place. Try something else and come back
later. Get the intuition of what is going on behind the few lines of code written to implement it.
With practice things will keep getting clearer. Keep reading about it from multiple sources.
4. Go through the quick learning videos before attending the class.
5. Make extensive notes when trying to understand / watching videos or attending the live
online class – it helps with internalizing the information as well as with the review.
6. Take a print out of the codes and presentations. During explanation, even if you are not
following completely due to the amount of information shared, making your own notes at that
time helps during review.
7. Understand limitations of each algorithm, if any.
8. Understand the usual application areas of each of the algorithms and why are they used
there.
9. Try understanding how these algorithms differ from each other. Using a single problem
statement and solving it using different applicable algorithms should help here.
10. Remember, the algorithms are just tools to solve problems. Don’t lose sight of the main
problem statement during implementation.
11. Many times, simple implementations are good enough. Build a simple solution first quickly
and then iterate - you may want to try different features, tuning the parameters and
hyperparameters, different algorithms, stacking different algorithms together and so on.
Make sure you try one thing at a time and not everything together since you would want to
know what change made the algorithm(s) better or worse.
12. Explain what you have done to one person who knows about the algorithms in technical
terms and to another who does not know the algorithms per se but can follow the problem
and its solution logically. Gaps in understanding are best understood when explaining to
others.
13. Learning is an iterative process – your first implementation may not be the best. It can be
made better over time. Please be patient.
14. Journal your progress. This may help you blog later.

How to ask good questions?

One of the most important ways to learn is to be able to ask questions well. Well written questions
encourage the community to answer them.
Kindly keep the following points in mind:
Give a context to the issue. Mention the problem you are trying to solve. Give a brief of all the steps
you have taken. Then mention where you may need assistance. Please mention all the things you
may have tried already. Then please be specific regarding the help you require. Do not give open
ended questions.
Examples of questions which do not help:
• What does question 5 mean in the assignment?

Examples of question which help us help you:


• I have been trying to improve the model performance for real estate to predict price. But I am
not able to go beyond the 0.47 threshold. I have imputed NA's with mean of non-missing
values for variables. Created dummy variables and dropped variables which are insignificant
to the model. Anything else that I am missing. I tried decision tree also, but still the
performance is 0.46.
When writing the question, imagine that the person reading it has no clue about the issue you are
facing. Help this person help you!

You might also like