A multi-manifold learning based instance weighting and under-sampling for imbalanced data classification problems

Journal of Big Data

Table 23 Average F-measure of the proposed method with three different sample weighting models

Dataset	Model 1	Model 2	Model 3
Dataset	weight = marg	weight = -cent	weight = marg-cent
ecoli1	0.85 ± 0.153	0.85 ± 0.153	0.85 ± 0.152
ecoli2	0.90 ± 0.105	0.90 ± 0.123	0.91 ± 0.105
ecoli3	0.86 ± 0.196	0.86 ± 0.196	0.86 ± 0.195
ecoli4	0.92 ± 0.171	0.92 ± 0.171	0.92 ± 0.170
ecoli0147vs56	0.89 ± 0.119	0.88 ± 0.141	0.89 ± 0.119
ecoli034_5	0.88 ± 0.151	0.88 ± 0.151	0.88 ± 0.151
ecoli0147_2356	0.81 ± 0.159	0.81 ± 0.159	0.81 ± 0.159
glass0	0.78 ± 0.076	0.78 ± 0.076	0.79 ± 0.116
glass0123456	0.92 ± 0.154	0.92 ± 0.154	0.92 ± 0.153
kddcup-buffer_overflow_vs_back	1.00 ± 0.000	1.00 ± 0.000	1 ± 0
new-thyroid1	0.93 ± 0.146	0.93 ± 0.146	0.98 ± 0.075
page-blocks-1-3_vs_4	0.96 ± 0.080	0.96 ± 0.080	096 ± 0.080
Pima	0.65 ± 0.071	0.65 ± 0.071	0.65 ± 0.067
segment0	0.98 ± 0.017	0.99 ± 0.018	0.98 ± 0.017
shuttle_2_vs_5	1.00 ± 0.000	1.00 ± 0.000	1 ± 0
vehicle2–1	0.98 ± 0.032	0.97 ± 0.033	0.97 ± 0.030
vowel0	0.90 ± 0.120	0.90 ± 0.120	0.90 ± 0.104
wisconsin	0.97 ± 0.050	0.97 ± 0.050	0.97 ± 0.050