1
$\begingroup$
  • Data Cleaning
  • Data Imbalance solving (Classification)
  • Data Smoothing (decreasing noise)
  • Creating-deleting features from original data
  • Data Transformation (Box-cox,Log Transform)
  • Making Dataset stationery (time series)
  • And other specific data preprocessing methods in NLP-Computer Vision (very specific ones)

I am trying to research data preparation methods and so far those are the things i could find. Do you think is anything missing? Thanks.

$\endgroup$
3
  • 1
    $\begingroup$ Your question is too broad. You are asking for preprocessing techniques for time series, computer vision, nlp, multiclass classification all in one question! I doubt anyone would answer you. Maybe try one question for separate topics $\endgroup$
    – spectre
    May 5 at 12:49
  • $\begingroup$ @spectre hello. Actually this is very good answer. And thank you for that. But the problem is, for those separate domains, there is always a missing data prep technique. Where can i find those domain-specific data preps. Methods in one place? Is there Any books or website that you can tell me? $\endgroup$
    – canP
    May 5 at 13:17
  • $\begingroup$ @spectre yes, i am asking every data prep. Method for every possible problem in the world. Thats correct and i think there should be an answer to that. If there is an answer to that in any resource i want to know it(as much as i can). $\endgroup$
    – canP
    May 5 at 14:09

1 Answer 1

0
$\begingroup$

Here are some.

  1. Dealing with the variable Types.
  2. Dealing with Missing data
  3. Encoding categorical variables
  4. Categorical variable — cardinality
  5. Categorical variable — rare labels
  6. Dealing with Outliers
  7. Variable Transformations
  8. Variable Discretization
  9. Feature Scaling and the list goes on.......

The question ased is too broad. These are some of the steps to be taken care of.

$\endgroup$
2
  • $\begingroup$ I mean are we sure that general (must be checked) general data prep. Methods are endless? The problem is whenever i look google it says : standarize, normalize, clean outlies, clean data. But my main problem in classification was "imbalanced dataset" and solving it costed me 27 hours. If there are endless data science problems in the world it means there are endless data prep method and this thing is impossible to handle. $\endgroup$
    – canP
    May 7 at 13:09
  • $\begingroup$ Imagine, you have time series data. And you are trying to find what is the problem. And boom. After 3 weeks you figured it out that problem was noise. But "noise" problem is one of the endless problems. So every single time for every single problem you have to waste at lease 90 hours. Sir i dont think this is correct. What do you think? $\endgroup$
    – canP
    May 7 at 13:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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