Predicting new instance with missing predictor I have a question about predicting a new instance with a missing predictor(s).
I am given a data. Let's say I preprocess, clean data and as a result, let's say, 10 predictors left. Then, I train my model on a resulting data, so I am ready to use model to predict.
Now, what should I do if I want to predict a new instance which 1 or 2 predictors are missing?
 A: Dealing with missing data in general is not easy, but provided the values are missing completely at random, multiple imputation is often used (see below).
Most importantly, you should consider the nature of missingness, among which three distinctions are usually made:


*

*Missing completely at random (MCAR): There is no reason to suspect that these values are missing because they are extreme or because of (an extreme value of) another variable;

*Missing at random (MAR): There is no reason to suspect that these values are missing because they are extreme, but they might be missing due to (an extreme value of) another variable;

*Missing not at random (MNAR): Data which is missing is missing because it is extreme. 


Naturally, MCAR is considered the least problematic and MNAR the most. When missing data is MCAR, imputation can be used to fill in the missing values. They can be replaced by the mean of non-missing values for example, or they can be imputed by means of regression.
See also this answer about missingness and this one about alternatives to simply imputating the mean.
