naiveIntegration: Combine probabilities from multiple data sources.
In pmartR/peppuR:

Combine probabilities from multiple data sources.

1	naiveIntegration(probabilities_by_source)

probabilities_by_source

a list of data.frames with one column for each predicted class corresponding to the probability of that class label given the data source. Each data frame corresponds to a data source. All probabilities in each row must sum to 1. The rows are generally cross validated test sets.

This function combines class probabilities from learning algorithms on multiple or single data sources over cross validation partitions. The mechanism for combining the probabilities is a posterior probability where

P(c_{j} | x_{i}) = /prod{P(c_{j} | x_{i}^{s})}_{s=1}^{ns}

where c is one of j class labels, x is observed data for one of ns data sources.

pmartR/peppuR documentation built on Jan. 17, 2020, 12:54 p.m.