R/rf_LUR.R

In mengluchu/APMtools: statistical air pollution mapping tools

#' Random forest for LUR (using ranger)

#' @param  variabledf the dataframe containing predictors and dependent variable
#' @param y_varname  name of the dependent variable.
#' @param training the index for the rows used for training.
#' @param test the index for the rows used for testing.
#' @param grepstring the variable/column names of predictors in Lasso, grepl stlye, e.g. 'ROAD|pop|temp|wind|Rsp|OMI|eleva|coast'
#' @return  plot variable importance and an error matrix
#' @export


rf_LUR = function(variabledf, vis = F, numtrees = 2000, mtry = 33, y_varname = c("day_value", "night_value", "value_mean"), training, test, grepstring, ...) {
    prenres = paste(y_varname, "|", grepstring, sep = "")
    pre_mat = subset_grep(variabledf[training, ], prenres)


    y_test = variabledf[test, y_varname]
    x_test = variabledf[test, ]

    formu = as.formula(paste(y_varname, "~.", sep = ""))

    rf3 <- ranger(formu, data = pre_mat, num.trees = numtrees, mtry = mtry, importance = "permutation")

    df = data.frame(imp_val = rf3$variable.importance)

    if (vis) {
        imp_plot = ggplot(df, aes(x = reorder(rownames(df), imp_val), y = imp_val, fill = imp_val)) + geom_bar(stat = "identity", position = "dodge") + coord_flip() + ylab("Variable Importance") + xlab("") + ggtitle(paste("Information Value Summary",
            y_varname, sep = ": ")) + guides(fill = F) + scale_fill_gradient(low = "red", high = "blue")

        print(imp_plot)
    }

    pre_rf <- predictions(predict(rf3, data = x_test))
    # rf_residual <- pre_rf - rdf_test$NO2
    return(error_matrix(y_test, pre_rf))


}