Dear Polash Banerjee

Please be specific about the environment you are using for establishing an ML model. Also, are you trying to do a classification or regression task? Based on your explanation I guess you are going to do a binary classification. Like Forest fire susceptibility map. For that, you need two types of data ie., forest fire occurrence and non-occurrence with x and y coordinates.

I guess you are going to use R software

1)Please load the corresponding package (like Caret, e1071 and so on)

2) Load your training data as CSV

3) train the ML model

4) Load all your independent variables ( rainfall, temperature, slope..... etc)

5) Use "prediction function" to predict the forest fire susceptibility based on your trained model.

6) Plot the predicted map and create a raster file

Here you have your final prediction map (raster)

Last but not least I doubt the credibility of the interpolation technique in such a way that you are described. Because in prediction mapping each cell is considered as separate values and the probability is calculated on a pixel to pixel basis. Hence interpolating the values for a prediction map will not be a good idea, I think.

regards

Thanks, Ashok for your reply. It's very naive of me to ask this question in this first place.

I forgot that the ML will generate the much-valued coefficient/ weights for the various input rasters. By using the appropriate hypothesis for all pixels I will be able to generate the desired spatial prediction model.

Thanks again!

Polash Banerjee

I had used some ML model in prediction mapping. Please feel free to ask questions if you need more explanations (R)

Best wishes

Thanks a lot. I will be in touch with you. Regards