ml_cor_filter: A function to filter models based on inter-model correlation.

In edwardcooper/automl: Train hundreds of machine learning models with a few lines of code.

Description

This function will also remove models that have NA value in resampled performance. a.k.a NA value in modelCor function output. Mainly based on modelCor function in caret.

Usage

ml_cor_filter(models, cor_level = 0.9)

Arguments

models	A list of caret models generated from ml_list function or by combining caret models into a list manually.
cor_level	A numeric, the maximum inter-model correlation computed from modelCor function from caret package.

Details

What the function does is to first get rid of model performance with missing values. Second, filter models based on cor_level, which is to get rid of the models with high correlation.

Value

A list of caret models that its performance metrics from cross-validation satisfy the conditions.

Examples

## Not run: 
 # a maximum of correlation 0.9 between models.
 low_cor_models_churn=testmodels_churn%>%ml_cor_filter(cor_level = 0.9)
 # a maximum of correlation 0.7 between models.
 low_cor_models_churn=testmodels_churn%>%ml_cor_filter(cor_level = 0.7)


## End(Not run)

edwardcooper/automl documentation built on June 3, 2019, 1:05 a.m.