Table 2 Core features of MILO classification algorithms.

From: Automated machine learning for endemic active tuberculosis prediction from multiplex serological data

 

Automated ML (MILO) approach

Algorithms

KNN, LR, SVM, DNN, RF, NB and GBM

Scaler(s) used

Standard scaler, min/max, and no scaler

Feature selector and/or transformers used

ANOVA F value select percentile (25% increments)

Random Forest Feature Importances (25% increments) and

Principal component analysis

Hyperparameter searchers

Grid search and

Random Search × 2

Scorer(s) used in the training/initial validation phase

Accuracy

ROC-AUC

F1

Model assessments

Generalization assessment on all pipelines

  1. Seven of the most widely validated and adopted algorithms (deep neural network (DNN), logistic regression (LR), Naïve Bayes (NB), k-nearest neighbors (k-NN), support vector machine (SVM), random forest (RF), and XGBoost gradient boosting machine (GBM)) are used in the pipelines generated in MILO’s automated approach, which also includes several hyperparameter tuning/search tools such as random search tool × 2 in addition to our custom grid search.
Back to article page