2
$\begingroup$

Suppose we have a data set of fields with negative integer values. So can we consider that fields with negative values for further proceedings or do we need to ignore that fields? If we can consider those negative values then please tell me how will we process it?

The dataset is bank_data

Improve this question
$\endgroup$
3
  • $\begingroup$ Why do you want to ignore negative fields? Does it impact the result? $\endgroup$
    – DuttaA
    Aug 22, 2018 at 7:15
  • $\begingroup$ Impossible to answer without knowing what the data means. If its invalid data then leave it out. If its valid then process as normal. Most functions will work perfectly well with negative, but again, depends entirely on what you are doing. $\endgroup$
    – TomC
    Aug 22, 2018 at 7:24
  • $\begingroup$ Depends on what negative values mean for that field. Like temperature can be negative also, you can't remove that. $\endgroup$
    – Ankit Seth
    Aug 22, 2018 at 7:43

2 Answers 2

Reset to default
1
$\begingroup$

I assume that you mean negative values that are not in the semantic domain of the feature and thus represent special cases that do not actually represent a value that is negative. If that assumption is correct, I'd suggest that you split the feature in two:

  1. A column representing the actual value - this would be blank/null for negative values; and
  2. A column that encodes the additional (possibly categorical) information, represented by the negative numbers in a better way. This could possibly be blank for those datapoints that have an actual value.

This, would force you to face the second problem you have: how to handle missing data.

Improve this answer
$\endgroup$
1
$\begingroup$

There are no problem with negative inputs values, as long as it's mean something. As Ankit Saith said in the comment, temperatures can be negative, money too (positive is money I earn, negative the money I lose) and so on. Of course, inputs like distances should not be negative !

Furthermore, generally in deep learning, you normalize your dataset to have inputs with 0 mean and a std of 1. Then you have "small" value around 0 which can be positive or negative !

Improve this answer
$\endgroup$

Your Answer

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

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