Are there any methods to deal with missing data?

Question

Machine learning

The bottom line is that there is a sample - 6 features, all real numbers. The problem of binary classification is being solved. But the peculiarity of the sample is that there are missing data. Not missing, but missing.
As an option, add binary features (there is data or not)
What other methods are there that allow you to work with this kind of data?

Answer 1

What does "Not missed, namely missing" mean. . Do I understand correctly that, for example, sign A for object X is present, but not for object Y? Well, then you need to choose the appropriate solution method. For example, for a classification problem, methods based on decision trees do not require the same set of features for all objects.

Answer 2

A head-on solution, since you have few features, search separately for all combinations of features present (that is, roughly speaking, if, for example, you have 6 features and 1 and 2 features may be missing in the data at the same time, then divide the sample into 20 parts for all possible combinations of missing features and for each look for your own classifier).
Sometimes, if it is possible to choose a value for features when their influence on the result is almost zero, and most importantly, these values ​​do not occur in the sample, then you can set such degenerate values ​​for unknown values ​​(for example, the value is the distance to the affected object, make it very large or vice versa zero ). Since you are classifying the data and not getting some value as a value, this can be justified.