0

So I have a database with 64954 observations of 37 variables, 16 of those are predictor variables. My independent variable "y" can be categorized into different "bins" depending of its value. For example if y<52% it belongs to the bin 1. If y > 52% it belongs to the bin 2, etc... I used Naive Bayes to predict the bin, given by certain values of the predictor variables.

But I'm curious to know if there's a way to "inverse" the problem, meaning if I give the program the bin that y belongs to, can it give me the values that the predictor variables most likely take? Thanks

Konrad Rudolph
  • 530,221
  • 131
  • 937
  • 1,214
Jia Hannah
  • 95
  • 7
  • The inverse problem has no longer anything to do with classification. You can consider all the observations in one bin as random sample from a population. Then it's "just" classical statistics: You can estimate the properties of the population using the sample. That could be as simple as computing the "mean sample" for each bin, but, depending on your covariance structure, it can also be much more difficult to compute the "most likely values" in a meaningful way. – AEF Aug 08 '22 at 12:55
  • This doesn't appear to be a specific programming question that's appropriate for Stack Overflow. If you are asking for statistical methods to do what you want, then you should seek help at [stats.se] instead. Otherwise questions here work best when you provide a [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with sample input and desired output that can be used to test and verify possible solutions. – MrFlick Aug 08 '22 at 13:30

0 Answers0