Module smart 

Smart Feature Engineering

Automatically infers optimal feature generation based on data characteristics and model analysis. Acts as a “Smart Feature Engineer” that prescribes features.

Design Philosophy

Different models benefit from different features:

• Linear models: Polynomial features (x², x³) help capture non-linearity
• Tree models: Interaction features (x_i * x_j) capture combinations trees struggle with
• LTT mode: Polynomial for linear phase, interactions for tree phase (on residuals)

Example

use treeboost::analysis::{DataFrameProfile, DatasetAnalysis};
use treeboost::features::smart::{SmartFeatureEngine, SmartFeatureConfig};

let profile = DataFrameProfile::analyze(&df, "target")?;
let analysis = DatasetAnalysis::analyze(&dataset);

let plan = SmartFeatureEngine::infer(&profile, Some(&analysis));
println!("Feature Plan:\n{}", SmartFeatureEngine::summarize(&plan));

Structs

FeaturePlan

Feature generation plan

LttFeaturePlan

LTT-specific feature plan with separate phases

SmartFeatureConfig

Configuration for smart feature engineering

SmartFeatureEngine

Smart Feature Engineering Engine

Enums

SmartFeaturePreset

Presets for smart feature engineering.

TimeFeatureType

Time feature types to extract from DateTime columns