Add interpretable SDM script

interpretable_sdm.r

@@ -0,0 +1,101 @@
+library(sdmbench)
+library(mlr)
+library(lime)
+library(dplyr)
+
+
+set.seed(42)
+
+# obtain occurence and environmental data
+occ_data_raw <- get_benchmarking_data("Loxodonta africana", limit = 1000)
+
+# rename environmental data features to be more understandable
+names(occ_data_raw$raster_data$climate_variables) <- c(
+  "mean_temp",
+  "mean_diurnal",
+  "isotherm",
+  "seas_temp",
+  "warmest_month",
+  "coldest_month",
+  "range_temp",
+  "wettest",
+  "driest",
+  "warmest_quart",
+  "coldest_quart",
+  "precip",
+  "precip_wettest_month",
+  "precip_driest_month",
+  "precip_season",
+  "precip_wettest_quart",
+  "precip_driest_quart",
+  "precip_warmest_quart",
+  "precip_coldest_quart"
+)
+
+names(occ_data_raw$df_data) <- c(
+  "mean_temp",
+  "mean_diurnal",
+  "isotherm",
+  "seas_temp",
+  "warmest_month",
+  "coldest_month",
+  "range_temp",
+  "wettest",
+  "driest",
+  "warmest_quart",
+  "coldest_quart",
+  "precip",
+  "precip_wettest_month",
+  "precip_driest_month",
+  "precip_season",
+  "precip_wettest_quart",
+  "precip_driest_quart",
+  "precip_warmest_quart",
+  "precip_coldest_quart",
+  "label"
+)
+
+# minor data processing
+occ_data <- occ_data_raw$df_data
+occ_data$label <- as.factor(occ_data$label)
+
+coordinates.df <- rbind(occ_data_raw$raster_data$coords_presence,
+                        occ_data_raw$raster_data$background)
+occ_data <- cbind(occ_data, coordinates.df)
+occ_data <- na.omit(occ_data)
+
+# split data into training and testing
+train_test_split <- rsample::initial_split(occ_data, prop = 0.7)
+data.train <- rsample::training(train_test_split)
+data.test  <- rsample::testing(train_test_split)
+
+train.coords <- dplyr::select(data.train, c("x", "y"))
+data.train$x <- NULL
+data.train$y <- NULL
+
+test.coords <- dplyr::select(data.test, c("x", "y"))
+data.test$x <- NULL
+data.test$y <- NULL
+
+# fit Random Forest
+task <- makeClassifTask(id = "model", data = data.train, target = "label")
+lrn <- makeLearner("classif.randomForest", predict.type = "prob")
+mod <- train(lrn, task)
+
+explainer <- lime(data.train, mod)
+explanation <- lime::explain(sample_n(data.test, 3), explainer, n_labels = 1, n_features = 12)
+plot_features(explanation)
+
+# create a custom function to make model object work with raster predictions
+customPredictFun <- function(model, data) {
+  v <- predict(model, data, type = "prob")
+  v <- as.data.frame(v)
+  colnames(v) <- c("absence", "presence")
+  return(v$presence)
+}
+
+# predict and create Habitat Suitability Map
+pr <- dismo::predict(occ_data_raw$raster_data$climate_variables, mlr::getLearnerModel(mod, TRUE), fun = customPredictFun)
+
+rf_map <- raster::spplot(pr, main = "Predicted Habitat Suitability Map (Random Forests)", ylab = "Suitability")
+