Hi Elizaveta Khazieva

Are you asking about the 'Maximum Likelihood Classification' function? I usually see it being used for predicted/supervised land classification.

Binary logistic regression is a statistical technique used to predict the relationship between predictors (your raster of independent variables) and a predicted variable (the dependent variable) where the dependent variable is binary (1/0).

Think of this as looking for the patterns and co-variations in the observed scores that tend to favor one outcome or the other. You can calibrate this model with your observations and obtain a predicted probability score for each case (even though the dependent variable was actually just one or zero). The trick is to use a transform like log (p/1-p). If this is regressed on a linear combination of your independent variables, you can easily back transform to get p. (p will be between 0 and 1)

See:

https://en.wikipedia.org/wiki/Logistic_regression

The Generalized Linear Regression tool is the function in Pro that will let you do a binary logistic regression. If your application would benefit more from a prediction, you might want to also take a look at the Forest Based Classification tool. I'm not sure if it will work with your training input (it uses vector training data), but it's the state-of the art for predictions.

https://pro.arcgis.com/en/pro-app/latest/get-started/create-points-on-a-map.htm

Just as an alternative idea, you can also consider using the raster calculator to aggregate values in different layers, then vectorizing and finally generating heatmaps (Gi*). The z-score (something similar to variance) can give you an idea of the occurrence of the phenomena in the study area.