Table 4 Model choice—Results of the 15 models with the 5-participant test set.

From: Classification of human walking context using a single-point accelerometer

 

Accuracy

F1-score

AUROC

Model S2

0.917

0.948

0.940

Model S3

0.917

0.948

0.935

Model S5

0.909

0.942

0.923

Model S6

0.920

0.950

0.921

Model S8

0.917

0.948

0.932

Model S9

0.929

0.956

0.918

Model S10

0.920

0.951

0.936

Model S11

0.917

0.948

0.924

Model S12

0.920

0.950

0.927

Model S13

0.926

0.954

0.926

Model S14

0.914

0.946

0.932

Model S15

0.932

0.958

0.917

Model S16

0.941

0.963

0.931

Model S18

0.923

0.952

0.930

Model S20

0.923

0.952

0.942

Best model

0.941

0.963

0.931

  1. Model Sx corresponds to the models trained and validated with the leave-one-participant-out scheme with Sx held out using the Random Forest algorithm with the biomechanics features set as shown in Table 3. Model S16 performs the best, as it had the highest accuracy, F1-score, and AUROC.
Back to article page