Skip to main content

Table 3 Model performance using MILR with smaller training sizes (the highest value for each metric is highlighted in bold type): the effect of formal reports was more obvious when the training size was smaller

From: Adverse event detection by integrating twitter data and VAERS

Twitter data

Formal

ACC

PR

RE

FS

AUC

#Training

#Report

     

314 (20%)

0

0.7731

0.7278

0.5923

0.6525

0.8446

 

500

0.7812

0.7323

0.6212

0.6713

0.8539

 

1000

0.8112

0.7993

0.6356

0.7076

0.8888

 

1500

0.8136

0.7935

0.6524

0.7151

0.8923

 

2000

0.8114

0.7812

0.6612

0.7156

0.8916

 

2500

0.8112

0.7824

0.6590

0.7147

0.8904

786 (50%)

0

0.7939

0.7689

0.6141

0.6816

0.8646

 

500

0.7920

0.7651

0.6125

0.6790

0.8684

 

1000

0.8041

0.7682

0.6567

0.7064

0.8834

 

1500

0.8034

0.7720

0.6482

0.7031

0.8834

 

2000

0.8092

0.7968

0.6312

0.7044

0.8897

 

2500

0.8066

0.7711

0.6615

0.7108

0.8866

1048 (67%)

0

0.7952

0.7841

0.5953

0.6767

0.8646

 

500

0.7850

0.7615

0.5915

0.6645

0.8653

 

1000

0.7983

0.7948

0.5937

0.6795

0.8843

 

1500

0.7996

0.7944

0.5992

0.6830

0.8880

 

2000

0.8034

0.7984

0.6080

0.6903

0.8899

 

2500

0.8060

0.8016

0.6133

0.6949

0.8910

1179 (75%)

0

0.7952

0.7845

0.5927

0.6752

0.8664

 

500

0.7933

0.7695

0.6010

0.6743

0.8846

 

1000

0.8034

0.7881

0.6172

0.6915

0.8948

 

1500

0.8041

0.7913

0.6154

0.6915

0.8963

 

2000

0.8041

0.7940

0.6119

0.6901

0.8983

 

2500

0.8041

0.7940

0.6119

0.6901

0.8985