Table 4 Performance of models -- Area under the ROC Curve (AUC)

original

0.798

0.823

0.829

0.753

0.767

0.776

0.670

0.674

0.662

0.912

MG

0.797

0.808

0.826

0.750

0.757

0.779

0.678

0.670

0.676

0.913

keywords selection (cut words: frequency < 2)

cubic interpolation

0.798

0.791

0.817

0.728

0.755

0.738

0.734

0.770

0.795

0.920

cut

0.803

0.831

0.837

0.723

0.778

0.782

0.671

0.675

0.658

0.922

mean

0.802

0.800

0.825

0.719

0.781

0.760

0.729

0.743

0.692

0.959

keywords selection (cut words: MI bottom 5%)

cubic interpolation

0.744

0.628

0.684

0.622

0.675

0.594

0.691

0.808

0.831

0.930

cut

0.712

0.641

0.754

0.689

0.753

0.716

0.653

0.616

0.551

0.929

mean

0.721

0.637

0.718

0.681

0.769

0.717

0.686

0.776

0.673

0.968

keywords selection (cut words: MI bottom 5% or frequency < 2)

cubic interpolation

0.776

0.645

0.689

0.623

0.683

0.742

0.688

0.798

0.841

0.928

cut

0.750

0.650

0.758

0.691

0.763

0.727

0.658

0.621

0.550

0.944

mean

0.754

0.644

0.725

0.677

0.775

0.696

0.688

0.792

0.628

0.965

keywords selection (cut words: frequency < 2) + MG

cubic interpolation

0.794

0.784

0.809

0.743

0.752

0.742

0.665

0.755

0.818

0.916

cut

0.794

0.807

0.845

0.764

0.759

0.786

0.682

0.670

0.675

0.904

mean

0.801

0.784

0.826

0.769

0.778

0.771

0.669

0.713

0.717

0.952

keywords selection (cut words: MI bottom 5%) + MG

cubic interpolation

0.681

0.626

0.662

0.643

0.659

0.591

0.726

0.803

0.836

0.933

cut

0.666

0.635

0.706

0.668

0.750

0.705

0.660

0.622

0.591

0.939

mean

0.693

0.640

0.690

0.660

0.771

0.690

0.671

0.702

0.797

0.959

keywords selection (cut words: MI bottom 5% or frequency < 2) + MG

cubic interpolation

0.709

0.638

0.661

0.642

0.663

0.720

0.705

0.793

0.845

0.919

cut

0.699

0.637

0.700

0.671

0.746

0.735

0.663

0.628

0.593

0.927

mean

0.678

0.657

0.683

0.677

0.765

0.698

0.687

0.698

0.766

0.951