Skip to main content

Table 10 Comparison of classification algorithm performance

From: Machine learning based GNSS signal classification and weighting scheme design in the built environment: a comparative experiment

Paper

Classification algorithm

QI

Classification accuracy (%)

Yozevitch et al. (2012)

Naïve threshold

C/N0

70–80

Hsu (2017)

SVM

C/N0

67.1

Hsu (2017)

SVM

Change rate of C/N0

39.4

Hsu (2017)

SVM

pseudorange residual

40.5

Hsu (2017)

SVM

difference between delta pseudorange and pseudorange rate

65.4

Sun et al. (2020a), (2020b)

GBDT

C/N0

74.1

Yozevitch et al. (2016)

Decision tree

C/N0, elevation angle, measurement, carrier lock, satellite clock bias, indifferent features

78.9

Hsu (2017)

SVM

C/N0, Change rate of C/N0, pseudorange residual, difference between delta pseudorange and pseudorange rate

75.4

Xu et al. (2019)

SVM

Correlator-Level and RINEX/NMEA-Level features

90.4

Sun et al. (2020a), (2020b)

GBDT

C/N0, pseudorange residual, elevation angle

89.0

Sun et al. (2020a), (2020b)

Decision tree

C/N0, pseudorange residual, elevation angle

76.0

Sun et al. (2020a), (2020b)

Distance-weighted KNN

C/N0, pseudorange residual, elevation angle

88.5

Sun et al. (2020a), (2020b)

ANFIS

C/N0, pseudorange residual, elevation angle

82.7

This paper, 2023

Random forest

The standard deviation of pseudorange, C/N0, elevation angle, and difference of azimuth angle

93.4