PFA landslide initiation model documentation

DESCRIPTION

@@ -1,27 +1,28 @@
-Package: TerrainWorksUtils
-Type: Package
-Title: General utility functions for TerrainWorks
-Version: 0.1.3
-Authors@R: 
-    c(
-        person("Helen", "Miller", email = "helenthehuman@gmail.com", role = c("aut", "cre")), 
-        person("Tate", "Brasel", role = "aut")
-     )
-Description: TerrainWorks resources for creating and assessing statistical models for predicting landslide susceptibility.
-License: MIT + file LICENSE
-Encoding: UTF-8
-LazyData: true
-Imports:
-  methods,
-  randomForest,
-  ROCR,
-  stats,
-  terra (>= 1.5)
-RoxygenNote: 7.2.3
-Suggests: 
-    knitr,
-    covr,
-    testthat (>= 3.0.0),
-    rmarkdown
-Config/testthat/edition: 3
-VignetteBuilder: knitr
+Package: TerrainWorksUtils
+Type: Package
+Title: General utility functions for TerrainWorks
+Version: 0.1.3
+Authors@R: 
+    c(
+        person("Helen", "Miller", email = "helenthehuman@gmail.com", role = c("aut", "cre")), 
+        person("Tate", "Brasel", role = "aut")
+     )
+Description: TerrainWorks resources for creating and assessing statistical models for predicting landslide susceptibility.
+License: MIT + file LICENSE
+Encoding: UTF-8
+LazyData: true
+Imports:
+  methods,
+  randomForest,
+  caret,
+  ROCR,
+  stats,
+  terra (>= 1.5)
+RoxygenNote: 7.2.3
+Suggests: 
+    knitr,
+    covr,
+    testthat (>= 3.0.0),
+    rmarkdown
+Config/testthat/edition: 3
+VignetteBuilder: knitr

R/build_models.R

@@ -33,7 +33,8 @@
 #' @return Nothing
 #' @export
 #'
-build_k_fold_rf_model <- function(data,
+build_k_fold_model <- function(data,
+                                  type,
                                   seed = NULL,
                                   ctrl_method = "repeatedcv",
                                   folds = 5,
@@ -44,6 +45,10 @@ build_k_fold_rf_model <- function(data,
     stop("Must provide data.")
   }
 
+  if (!type %in% c("rf", "glm")) {
+    stop("Must specify model type. Current options are: rf, glm")
+  }
+
   if (!is.data.frame(data) | is.null(data$class)) {
     stop("Data must be a data frame (or coercible) with a \"class\" column")
   }
@@ -68,7 +73,7 @@ build_k_fold_rf_model <- function(data,
                                      data = data,
                                      preProcess = preprocess,
                                      trControl = ctrl,
-                                     method = "rf",
+                                     method = type,
                                      metric = "AUC"))
 
   print(model)
@@ -80,7 +85,6 @@ build_k_fold_rf_model <- function(data,
 }
 
 
-
 #' @title Build Random Forest Model
 #'
 #' @param data A dataframe containing columns with model inputs and a

R/create_model_dataframe.R

@@ -0,0 +1,35 @@
+# define where to find the data and load it.
+folder <- "//Mac/Home/Documents/terrainworks/code/sandbox/data/downloaded_3.6/out/"
+folder <- "/Users/julialober/Documents/terrainworks/code/sandbox/data/downloaded_3.6/out/"
+load(paste0(folder, "ls_1996.Rdata"))
+load(paste0(folder, "ls_2007.Rdata"))
+load(paste0(folder, "ls_2011.Rdata"))
+load(paste0(folder, "nonls_1996.Rdata"))
+load(paste0(folder, "nonls_2007.Rdata"))
+load(paste0(folder, "nonls_2011.Rdata"))
+
+# remove these lines when the bug is fixed. the $geo field in the 2011 and 2007
+# data frames is not stored as an integer value
+ls_2011$geo <- ls_2011$geo$GeoClass
+ls_2007$geo <- ls_2007$geo$GeoClass
+
+# add year labels to each selection of landslide points and non-ls points
+ls_1996$year <- rep(1996, length(ls_1996[, 1]))
+ls_2007$year <- rep(2007, length(ls_2007[, 1]))
+ls_2011$year <- rep(2011, length(ls_2011[, 1]))
+nonls_1996$year <- rep(1996, length(nonls_1996[, 1]))
+nonls_2007$year <- rep(2007, length(nonls_2007[, 1]))
+nonls_2011$year <- rep(2011, length(nonls_2011[, 1]))
+
+# combine into one data frame.
+subset96 <- seq(1, length(nonls_1996[,1]), 10) # 1:length(ls_1996[, 1])
+subset07 <- seq(1, length(nonls_2007[,1]), 10) # 1:length(ls_2007[, 1])
+subset11 <- seq(1, length(nonls_2011[,1]), 10) # 1:length(ls_2011[, 1])
+training_data <- rbind(ls_1996,
+                       ls_2007,
+                       ls_2011,
+                       nonls_1996[subset96, ],
+                       nonls_2007[subset07, ],
+                       nonls_2011[subset11, ])
+
+training_data$class <- as.factor(training_data$class)