A Classification and Novel Class Detection Algorithm for Concept Drift Data Stream Based on the Cohesiveness and Separation Index of Mahalanobis Distance

Table 4

Experimental result data in KDD Cup dataset.

C	C&NCBM accuracy (%)	MineClass accuracy (%)	KNN accuracy (%)	C&NCBM evaluation time (s)	MineClass evaluation time (s)	KNN evaluation time (s)
1	99.8227	99.6988	99.6174	26.8109	25.95	22.4859
2	99.8468	99.7375	99.5071	54.1313	51.12	45.0750
3	99.8078	99.6569	99.4002	83.9797	77.80	69.3438
4	99.8541	99.7427	99.5081	117.6719	109.33	97.3688
5	99.8819	99.7942	99.6065	152.3984	138.35	123.5469
6	99.8890	99.8130	99.6721	190.0578	170.77	154.1313
7	99.8950	99.8156	99.6966	213.7469	203.87	185.8859
8	99.9028	99.8290	99.7249	246.0797	228.09	207.2969
9	99.9020	99.8245	99.7101	259.0313	258.43	234.4469