1
$\begingroup$

Having been in industry for a while, my puzzle on this question still remains unsolved. What exactly is the relationship between visualization and feature engineering? While on Kaggle or elsewhere we can frequently bump into beautiful visualization of data. However, I myself see it little reference values when coming into how it helps in feature engineering. In particular, I can complete the entire feature engineering without plotting any graphs and still manage to get highly accurate models, by simply relying on statistics I print from data. And more often than not, those graphs neither fail to show precisely the intervals or the exact numerical values for a certain data field. To me, it's too approximative to be useful.

Thus, I'm seeing there's little use of visualization if I already can feature engineer based on calculating some statistics. I am looking for someone to correct my concept and point out what I've been missing in terms of the contributions of visualization in feature engineering, so I won't be so blind. TIA.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

First of all you are nothing you mentioned in your question is that serious! You can do many things without visualisation and if you see a blank page of thousands of numbers more intuitive than a simple plot then lucky you! Nothing is right or wrong here!

Second point I would like to mention is that you can intuitively understand the distribution of your data in terms of classes or clusters in a simple visualisation that I am sure even you, yourself, will not see easily in just plain numbers. That was just an example and follows the rule of "an image worth thousand words"!

PS: Can you really detect skewness of distributions just from looking at numbers?!

Improve this answer
$\endgroup$
1
  • $\begingroup$ thanks for the opinion. well, indeed in some case looking at the numbers can't tell. Just that in MOST of the cases, i see people visualize for the sake of it without even touching or leveraging in any ways the visualization results into feature engineering. That't the part puzzling me. For example, when they find out which type of products are most popular in a period of time, and the analysis stops short there. FE part is not leveraging this fact. Thus, the motivation of this question: seeking the connections between the two. $\endgroup$
    – Student
    Commented Apr 22, 2022 at 0:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.