Dummy / Regressor Layer

Dummy Regressor: A baseline regression model using simple rules.

Purpose:

Provides regression baseline
Validates more complex models
Sanity checks performance
Tests pipeline infrastructure

Strategies:

Mean prediction
Median prediction
Quantile-based prediction
Constant value prediction

Use cases:

Model comparison baseline
Pipeline debugging
Null hypothesis testing
Data quality assessment

Outputs:

Predicted Table: Results with predictions
Validation Results: Cross-validation metrics
Test Metric: Hold-out performance
Feature Importances: Always zero (dummy model)

Note: Ignores feature values in predictions

Table

Predicted Table

Validation Results

Test Metric

Feature Importances

SelectTarget

column

Target column specification for regression:

Requirements:

Data type:
- Numeric values only
- Continuous or discrete
- No missing values
- Real-valued targets
Quality checks:
- Value range verification
- Distribution assessment
- Outlier detection
- Scale consideration
Statistical properties:
- Central tendency
- Spread measures
- Distribution shape
- Outlier impact

Note: Must be a single numeric column

Strategy

enum

Quantile

Prediction strategy for baseline regression:

Selection criteria:

Data distribution
Target characteristics
Baseline requirements
Performance comparison needs

Usage:

Model validation
Performance benchmarking
Statistical testing
Pipeline verification

Mean ~

Mean-based prediction strategy:

Formula: $\overset{y}{^} = \frac{1}{n} i = 1 \sum n y_{i}$

Characteristics:

Minimizes MSE
Central tendency measure
Sensitive to outliers
Simple computation

Best for:

Normal distributions
MSE optimization
General baselines
Initial benchmarks

Median ~

Median-based prediction strategy:

Formula: Middle value of sorted {y₁, ..., yₙ}

Properties:

Robust to outliers
Minimizes MAE
Central position measure
Order statistic

Best for:

Skewed distributions
Outlier presence
MAE optimization
Robust baselines

Quantile ~

Quantile-based prediction strategy:

Formula: q-th quantile of {y₁, ..., yₙ}

Features:

User-specified quantile
Distribution percentiles
Flexible prediction point
Risk-level modeling

Best for:

Quantile regression
Risk assessment
Asymmetric losses
Specific percentiles

Constant ~

Constant value prediction strategy:

Method: $\overset{y}{^} = c$ (user-defined constant)

Applications:

Fixed value prediction
Known target level
Specific baseline
Domain knowledge

Best for:

Known reference points
Simple comparisons
Specific hypotheses
Custom baselines

Quantile

f64

0.5

The quantile to predict using the “quantile” strategy. A quantile of 0.5 corresponds to the median, while 0.0 to the minimum and 1.0 to the maximum.

Range: [0.0, 1.0] where:

0.0: Minimum value
0.5: Median (default)
1.0: Maximum value

Common values:

0.25: First quartile
0.75: Third quartile
0.95: 95th percentile

Use cases:

Risk assessment
Confidence bounds
Distribution analysis
Extreme value study

ConstantValue

f64

Fixed prediction value using Constant strategy:

Usage:

Domain-specific constant
Known reference point
Theoretical value
Baseline comparison

Applications:

Null hypothesis testing
Known standards
Specific benchmarks
Reference comparisons