I want to use latitude and longitude as a feature for models like SVM or Logistic Regression (both for classification). What is the most common approach to use latitude and longitude values as features? I have tried to use decimal representation, I've also tried to scale it (for SVM) and the results were the same as without the latitude/longitude. These features are supposed to be significant in the performance of the model (according to my specific dataset of course).
I think representation of the data is less of a problem here. Although coordinate data are frequently used as predictors, the problem with those data is that they are autocorrelated eg Why does including latitude and longitude in a GAM account for spatial autocorrelation? So the tricky part is correctly modeling this variance structure, a construct of which can then be used as a predictor. One possibility is Moran Eigenvector Maps (MEM). On a separate point, depending on your dataset size (if it is not extremely large), Random Forests may be a more powerful technique than SVM.