Content Comparison

library(ForestForesight)
ff_environment() #after running this line, fill in the details in the interactive window below in Rstudio
#fastest is to select a small country with fewer tiles, e.g. 
#Suriname (SUR) has one, 
#Laos (LAO) two, 
#while Brazil has over fifteen tiles.

#first download the preprocessed data
ff_sync(features = "high", #we only want the most important features
        date_start = "2023-01-01", date_end = "2023-12-01",# this is an example year
        download_model = TRUE,  #we will download the already pregenerated model
        download_predictions = TRUE, #we will download the predictions as well
        download_groundtruth = TRUE #we will also download the groundtruth data since we have no own groundtruth
        )

# then use ff_run to load the input data, train and predict. 
# This will automatically plot the accuracy over the four prediction months per 1x1 degree tile
# and per month and the importance of individual features for the model. If not, increase the window
# size in the bottom right and run again.
ff_output <- ff_run(train_dates = ForestForesight::daterange("2023-01-01","2023-03-01"), #train on the first 3 months, normally we recommend at least a year of training data
                    prediction_dates = ForestForesight::daterange("2023-09-01","2023-12-01")) # predict on the last four months to always have a 6 month gap between trianing and predicting

# the plot below will show the four rasters for the four months of prediction
plot(ff_output$predictions)
# the plot below shows the medium, high and very high risk level polygons for december 2023
plot(ff_output$shape)
plot(ff_output$risk_zones$`2023-12-01`$medium, col="yellow", border = "yellow", add=TRUE)
plot(ff_output$risk_zones$`2023-12-01`$high, col="orange", border = "orange", add=TRUE)
plot(ff_output$risk_zones$`2023-12-01`$very_high, col="red", border = "red", add=TRUE)

#this shows the first lines of the accuracy dataframe, with precision, recall and F0.5 for every month
# and every 1x1 degree
head(ff_output$accuracy_dataframe)

#this shows the relative importance of the features for the built model
head(ff_output$importance_dataframe,10)

#in this variable the model is stored. You can save it by setting the model_save_path in ff_run before running.
ff_output$model

# by setting the parameters for output in ff_run you can export all this information to disk as well