50 likes | 170 Views
This document explores a pruning technique applied to outlier detection, specifically using a distance-based method where k=1 for nearest neighbors. It illustrates cases of outlier identification, demonstrating the best and worst scenarios in terms of total comparisons. Note the differences in total comparisons based on the ordering of outliers, revealing insights into the efficiency of the technique. Overall, this analysis emphasizes how pruning can enhance the performance of outlier detection algorithms by reducing unnecessary computations.
E N D
2 1 1 Example • Using the distance based method where knn=1, the left-most point is outlier • Using pruning: • What is the best case? • What is the worst case?
Outliers has first ordering 2 1 1 1 2 3 4 • Total # comparisons = 10
Outliers has last ordering 2 1 1 4 3 2 1 • Total # comparisons = 12
Another case 2 1 1 2 3 1 4 • Total # comparisons = 8