Skip to main content

Table 6 Summary of accuracies for module 3 with best classifiers and parameters for our 10-feature set, this time trained without gender

From: Sparsifying machine learning models identify stable subsets of predictive features for behavioral detection of autism

Module

3

3

Number of features

9

9

Best classifier

L2 LR

L1 Lin SVM

Optimal parameters

C = 1

C = 0.5

Area under ROC

0.95

0.95

Precision

0.99

0.99

Recall/sensitivity

0.89

0.95

Specificity

0.90

0.87

Balanced accuracy

0.90

0.91

F1 score

0.94

0.97

  1. LR denotes logistic regression, and L1 Lin SVM denotes L 1-penalized linear SVM
Back to article page

Molecular Autism

ISSN: 2040-2392