ensemble.model: Ensemble model (pre-trained ensemble classifier)

In haukelicht/politicaltweets: Classify political tweets

Ensemble model (pre-trained ensemble classifier)

Description

Pre-trained ensemble classifier blended with metric = "PR-AUC" constituent.models using caretEnsemble.

Usage

ensemble.model

Format

A 'caretEnsemble' object. Element 'models' is identical to constituent.models Element 'ens_model' is the ensemble classifier induced by PR-AUC-based blending.

Details

Individual models were tuned to individually defined tuning grids using 5-times repeated 10-fold cross validation to 6292 labeled tweets. "Best" models were selected among these candidates based on which tuning parameter combinations yielded the highest F1 value. These "best" candidates were then trained to the complete training data in a 10-times repeated 10-fold cross validation scheme.

The ensemble classifier was then obtained using caretEnsemble to blend them according to their test set performance (measured by PR-AUC) in these 100 resamples.

haukelicht/politicaltweets documentation built on July 3, 2023, 4:11 a.m.