Unified Interface for Machine Learning Models

Description

Provides a unified R6-based interface for various machine learning models with automatic interface detection, consistent cross-validation, model interpretations via numerical derivatives, and visualization. Supports both regression and classification tasks with any model function that follows R's standard modeling conventions (formula or matrix interface).

Package Content

Index of help topics:

Model                   Unified Machine Learning Interface using R6
cross_val_score         Cross-Validation for Model Objects
rcpp_hello_world        Simple function using Rcpp
unifiedml-package       Unified Interface for Machine Learning Models

Maintainer

T. Moudiki <thierry.moudiki@gmail.com>

Author(s)

T. Moudiki [aut, cre]

Unified Machine Learning Interface using R6

Description

Provides a consistent interface for various machine learning models in R, with automatic detection of formula vs matrix interfaces, built-in cross-validation, model interpretability, and visualization.

An R6 class that provides a unified interface for regression and classification models with automatic interface detection, cross-validation, and interpretability features. The task type (regression vs classification) is automatically detected from the response variable type.

Public fields

Methods

Public methods

• Model$new()

• Model$fit()

• Model$predict()

• Model$print()

• Model$summary()

• Model$plot()

• Model$clone_model()

• Model$clone()

Method new()

Initialize a new Model

Usage

Model$new(model_fn)

Arguments

Returns

A new Model object

Method fit()

Fit the model to training data

Automatically detects task type (regression vs classification) based on the type of the response variable y. Numeric y -> regression, factor y -> classification.

Usage

Model$fit(X, y, ...)

Arguments

Returns

self (invisible) for method chaining

Method predict()

Generate predictions from fitted model

Usage

Model$predict(X, type = NULL, ...)

Arguments

Returns

Vector of predictions

Method print()

Print model information

Usage

Model$print()

Returns

self (invisible) for method chaining

Method summary()

Compute numerical derivatives and statistical significance

Uses finite differences to compute approximate partial derivatives for each feature, providing model-agnostic interpretability.

Usage

Model$summary(h = 0.01, alpha = 0.05)

Arguments

Details

The method computes numerical derivatives using central differences.

Statistical significance is assessed using t-tests on the derivative estimates across samples.

Returns

A data.frame with derivative statistics (invisible)

Method plot()

Create partial dependence plot for a feature

Visualizes the relationship between a feature and the predicted outcome while holding other features at their mean values.

Usage

Model$plot(feature = 1, n_points = 100)

Arguments

Returns

self (invisible) for method chaining

Method clone_model()

Create a deep copy of the model

Useful for cross-validation and parallel processing where multiple independent model instances are needed.

Usage

Model$clone_model()

Returns

A new Model object with same configuration

Method clone()

The objects of this class are cloneable with this method.

Usage

Model$clone(deep = FALSE)

Arguments

Author(s)

Your Name

Examples

# Regression example with glmnet
library(glmnet)
X <- matrix(rnorm(100), ncol = 4)
y <- 2*X[,1] - 1.5*X[,2] + rnorm(25)  # numeric -> regression

mod <- Model$new(glmnet::glmnet)
mod$fit(X, y, alpha = 0, lambda = 0.1)
mod$summary()
predictions <- mod$predict(X)

# Classification example  
data(iris)
iris_binary <- iris[iris$Species %in% c("setosa", "versicolor"), ]
X_class <- as.matrix(iris_binary[, 1:4])
y_class <- iris_binary$Species  # factor -> classification

mod2 <- Model$new(e1071::svm)
mod2$fit(X_class, y_class, kernel = "radial")
mod2$summary()

# Cross-validation
cv_scores <- cross_val_score(mod, X, y, cv = 5)

Cross-Validation for Model Objects

Description

Perform k-fold cross-validation with consistent scoring metrics across different model types. The scoring metric is automatically selected based on the detected task type.

Usage

cross_val_score(
  model,
  X,
  y,
  cv = 5,
  scoring = NULL,
  show_progress = TRUE,
  cl = NULL,
  ...
)

Arguments

Value

Vector of cross-validation scores for each fold

Examples

library(glmnet)
X <- matrix(rnorm(100), ncol = 4)
y <- 2*X[,1] - 1.5*X[,2] + rnorm(25)  # numeric -> regression

mod <- Model$new(glmnet::glmnet)
mod$fit(X, y, alpha = 0, lambda = 0.1)
cv_scores <- cross_val_score(mod, X, y, cv = 5)  # auto-uses RMSE
mean(cv_scores)  # Average RMSE

# Classification with accuracy scoring
data(iris)
X_class <- as.matrix(iris[, 1:4])
y_class <- iris$Species  # factor -> classification

mod2 <- Model$new(e1071::svm)
cv_scores2 <- cross_val_score(mod2, X_class, y_class, cv = 5)  # auto-uses accuracy
mean(cv_scores2)  # Average accuracy

Simple function using Rcpp

Description

Simple function using Rcpp

Usage

rcpp_hello_world()	

Examples

## Not run: 
rcpp_hello_world()

## End(Not run)