train_model: Train various predictive models

In konradsemsch/aider: Perform everyday analytical and data science tasks more efficiently

Description

This function automatically builds different predictive models with reasonable default settings based on the implementation in the caret package. Data preprocessing can happen automatically through applying aider's default recipe blueprint. It is currently only implemented for classification problems. The default resampling procedure is repeated cross-validation.

Usage

train_model(df, target, type = "classification", models = c("rf"),
  use_recipe = TRUE, folds = 5, repeats = 5, upsample = "yes")

Arguments

df	A data frame
target	A target variable
type	Specify the modelling task. Possible options are: "classification" (default) and "regression"
models	Specify type of models to train. Possibile options are: "rf" (Random Forest) as default, as well as "en" (Elastic-Net), "svm" (Support Vector Machines) and "xgb" (XgBoost)
use_recipe	Specify whether a standardized recipe should be applied. If FALSE then the dataset needs to pre-processed before applying the function. Defaults to TRUE
folds	Specify the number of folds in cross-validation. Defaults to 5
repeats	Specify the number of times the fitting process should be repeated. Defaults to 5
upsample	Should the minority class be upsampled during resampling? Defaults to "yes"

Examples

data <- recipes::credit_data %>%
  first_to_lower()

models <- train_model(data, status, repeats = 1)

konradsemsch/aider documentation built on May 22, 2019, 2:40 p.m.