Table 7 Performance comparison with statistical analysis.

From: Intelligent ship traffic supervision system based on distributed blockchain and federated reinforcement learning for collaborative decision optimization

Method

Accuracy (%)

95% CI

Response Time (ms)

95% CI

Throughput (TPS)

Energy (kWh/1000tx)

Comm. Overhead (MB/decision)

p-value*

Traditional centralized

78.5 ± 2.1

[76.4, 80.6]

1,250 ± 89

[1,161, 1,339]

45 ± 5

2.4 ± 0.3

0.8 ± 0.1

-

Pure blockchain

82.1 ± 1.8

[80.3, 83.9]

980 ± 67

[913, 1,047]

120 ± 12

1.8 ± 0.2

1.2 ± 0.2

0.031

Pure federated learning

84.7 ± 2.3

[82.4, 87.0]

850 ± 54

[796, 904]

95 ± 8

0.9 ± 0.1

2.1 ± 0.3

0.015

Hybrid methods

88.3 ± 1.9

[86.4, 90.2]

720 ± 43

[677, 763]

180 ± 15

1.5 ± 0.2

1.6 ± 0.2

0.008

Proposed method

93.6 ± 1.4

[92.2, 95.0]

520 ± 31

[489, 551]

285 ± 18

0.12 ± 0.02

1.8 ± 0.1

< 0.001

  1. *p-values from t-tests comparing with traditional approach. Results show statistically significant improvements (p < 0.05).
Back to article page