Does someone have an elementary example (data sets and script) which could explain in which way remote sensing data should be converted to make k-NN estimation of forest parameters using R software. If it is possible please provide a link
Article yaImpute: An R Package for kNN Imputation
Hi Stefan,
thank you for response, I have already used these data from http://www.jstatsoft.org/v23/i10. As I understood authors used kNN=1 in this example. Maybe you comment me how to find an optimal number of nearest neighbour using yaImpute.
Looking forward to it...
Hi Victor,
Generally there is no thumb rule for selecting the value. If you are not sure about the parameter selection, you can run the prediction model for several values (for example kNN=1,2,3,...20) of kNN and analyze the prediction accuracy (R^2 and RMSE). For a more robust prediction model, you can add a cross-validation step.
-Ani
I need to do a cross-validaton and I am using yaImpute package. Have you any idea?
-Carmen
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing