Table 11 Ablation study on banana and collision datasets. Each dataset reports accuracy, reduction ratio, computational cost, and a task specific efficiency metric (SVs reduction for banana, carbon emissions for collision).

From: KM-DBSCAN: an enhanced density and centroid based border detection framework for data reduction towards green AI

Method

Accuracy

Reduction ratio

Task specific metric

Time (s)

Speedup

Banana

 Whole dataset

0.9044

–

–

0.0982

–

 K-Means only

0.9025

0.107

0.257 (SVs)

0.0031

31.67

 DBSCAN only

0.8755

0.255

0.311 (SVs)

0.0122

8.04

 KM-DBSCAN

0.9107

0.0372

0.0544 (SVs)

0.0010

98.2

Collision

 Whole dataset

0.9137

–

1.500 (g CO\(_2\))

853.58

–

 K-Means only

0.8861

0.099

0.1654 (g CO\(_2\))

81.888

10.423

 DBSCAN only

0.7372

0.110

0.1797 (g CO\(_2\))

96.437

8.85

 KM-DBSCAN

0.8932

0.079

0.1328 (g CO\(_2\))

72.4598

11.780

Back to article page