Table 5 The performance of different methods when the training data and validation data come from the same building
From: A universal Wi-Fi fingerprint localization method based on machine learning and sample differences
Algorithm | Building0 | Building1 | Building2 | ||||||
|---|---|---|---|---|---|---|---|---|---|
Floor accuracy (%) | Point accuracy (%) | Mean error (m) | Floor accuracy (%) | Point accuracy (%) | Mean error (m) | Floor accuracy (%) | Point accuracy (%) | Mean error (m) | |
NN | 92.35 | 71.54 | 4.063 | 97.41 | 82.68 | 3.278 | 79.49 | 63.23 | 13.098 |
DT | 37.89 | 2.73 | 23.602 | 46.43 | 6.12% | 24.317 | 33.25 | 1.29 | 39.034 |
RF | 98.44 | 71.88 | 2.102 | 97.96 | 73.98 | 3.050 | 96.39 | 74.23 | 1.880 |
Bagging | 95.70 | 61.33 | 3.611 | 95.41 | 63.78 | 4.403 | 95.62 | 63.40 | 3.488 |
XGBoost | 99.22 | 66.41 | 2.414 | 98.98 | 65.31 | 4.269 | 100.00 | 62.11 | 3.563 |
GBDT | 100.00 | 78.52 | 1.399 | 98.98 | 76.53 | 2.873 | 98.45 | 74.49 | 3.089 |