Table 3 Impact of predictive properties on execution times when de-identifying high-dimensional data. The figure shows the performance achieved with the BFS algorithm when using or not using predictive properties
From: Efficient and effective pruning strategies for health data de-identification
Attributes | 4 | 5 | 6 | 7 | 8 | 9 |
Without prediction [s] | 0.021 | 0.035 | 0.071 | 0.229 | 0.545 | 4.47 |
With prediction [s] | 0.019 | 0.033 | 0.067 | 0.223 | 0.533 | 4.13 |
Improvement | 9.52 % | 5.71 % | 5.63 % | 2.62 % | 2.20 % | 7.68 % |
Attributes | 10 | 11 | 12 | 13 | 14 | 15 |
Without prediction [s] | 24.02 | 70.19 | 215.55 | 1032.19 | 2811.32 | 8259.83 |
With prediction [s] | 21.69 | 61.24 | 185.51 | 858.41 | 2248.37 | 6843.33 |
Improvement | 9.72 % | 12.75 % | 13.93 % | 16.83 % | 20.02 % | 17.15 % |