I have two class of data. I want to separate their features that is strong wind and weak wind and I have two variables and a specific threshold that can be used to separate the features but I want to train a model and want that if the data changes with time, the threshold for wind to separate also should adapt accordingly.
Can anyone suggest on this which machine learning model should be more focused and how the dynamic threshold can be prepared.

I tried training ML model based on logistic regression using all features such as mean median skewness and kurtosis.
But the model fail to generate accurately