Table 8 Performance comparison (in terms of classification accuracies) of our proposed wrapper-based deep feature optimization method with some existing HAR works implemented on the HARTH, KU-HAR, and HuGaDB datasets.

From: Wrapper-based deep feature optimization for activity recognition in the wearable sensor networks of healthcare systems

Dataset

Researcher

Accuracy(%)

HARTH

Logacjov et al.14

81

Proposed wrapper-based deep feature optimization HAR framework

88.89

KU-HAR

Sikder and Nahid15

89.67

Proposed wrapper-based deep feature optimization HAR framework

97.86

HuGaDB

Gochoo et al.36

92.5

Logacjov et al.14

81

Filtjens et al.41

83.8

Javeed et al.37

92.5

Fang et al.38

79.25

Sun et al.39

88

Kumari et al.40

91.1

Proposed wrapper-based deep feature optimization HAR framework

93.82

  1. Bold indicates the least number of features selected and highest values of accuracy, Recall, Precision and F1-score attained for each of the HAR datasets.