Stochastic modeling poses still a real challenge in many areas. The meteorology domain is not an exception. Several related physical variables (e.g. temperature, pressure, humidity or wind) affect the modeling of precipitation activity. Modeling precipitation estimation accurately is thus a highly non-trivial task. In this paper we use regression with Random forest to model precipitation estimation. We use radar data in the form of RGB raster images. These images are publicly available from the Slovak Hydrometeorological Institute (SHMÚ) online 24/7. However, such representation of radar data introduces some degradation that negatively impacts the resulting model’s accuracy. We thus propose a compensation – bilinear interpolation on the raster images to elevate accuracy and reduce noise. We also propose additional attributes (including the presence of the cloud) to further improve the accuracy. In the result we were able to improve the accuracy of our model to 0.8479 with relatively low mean average error 0.0503.