Feature Engineering Agent • LLMAgentR

What This Agent Does

build_feature_engineering_agent() creates/derives features to improve model performance and explains the transformation logic.

Workflow Diagram

Generate Mermaid PNGs

library(LLMAgentR)

my_llm_wrapper <- function(prompt, verbose = FALSE) "LLM response placeholder"

workflow <- build_feature_engineering_agent(
  model = my_llm_wrapper,
  output = "both",
  direction = "LR"
)

save_mermaid_png(
  x = workflow,
  file = "pkgdown/assets/feature-engineering-agent-workflow.png"
)

Step 1: Build the Agent

library(LLMAgentR)

my_llm_wrapper <- function(prompt, verbose = FALSE) "LLM response placeholder"

feature_agent <- build_feature_engineering_agent(
  model = my_llm_wrapper,
  human_validation = FALSE,
  bypass_recommended_steps = FALSE,
  bypass_explain_code = FALSE,
  verbose = FALSE
)

Step 2: Run with State

data <- transform(mtcars, target = ifelse(mpg > median(mpg), "high", "low"))

initial_state <- list(
  data_raw = data,
  target_variable = "target",
  user_instructions = "Create useful derived numeric and categorical features.",
  max_retries = 3,
  retry_count = 0
)

final_state <- feature_agent(initial_state)
str(final_state)

Notes for Beginners

Always set target_variable for supervised tasks.
Review engineered features for leakage before training models.
Keep human validation on for high-stakes use cases.