Application of Machine Learning and Deep Learning in diagnosis Covid19

In hospitalized COVID-19 patients from Algerian (80) and French (99) cohorts, logistic regression with feature selection reached test accuracies up to 0.95 across death, ICU, and clinical status tasks.

• Why it matters: Limited healthcare resources during the pandemic increased the need for models that can forecast key clinical outcomes from patient data, while addressing a gap in reliable prediction under small datasets.
• What they did: The study evaluated five ML classifiers (LR, KNN, CART, NB, SVM) with ANOVA, MI, and RFE feature selection, using 30 repetitions of five-fold cross-validation plus a separate hold-out test set; MLPs with one to five hidden layers were also compared.
• The result: Best classifier–feature-selection pairings varied by dataset–outcome task, and ML test-set accuracies ranged from 0.47 to 0.95, exceeding average MLP accuracy in five of six tasks, but results require validation on larger independent datasets.

AI summary of the abstract · may contain errors · not medical advice · read the paper.