Random Forest prediction models for flowering in Southeast Asia

Finally, I wanted to see how well the climate and El Niño indicators can be used to predict the flowering in Southeast Asia. I ran a Random Forest analysis on the 5 regions. Table 5 below shows the prediction output. The model predicts the years when no flowering occurs fairly well. However the predictions for mass flowering years are far less accurate.  The best prediction of flowering is in North Peninsular Malaysia with 66 % accuracy (100% -class error 44%). In general the model also does a better job predicting strong flowering events rather than the weaker ones. For predicting phenology, accuracy of 50% is a good start.

                              Table 5: Output of random forest flowering predictions using climate and El Niño variables for the 5 regions through out Southeast Asia