Classifying microscopic images as acute lymphoblastic leukemia by Resnet ensemble model and Taguchi method

BMC Bioinformatics

Table 5 Accuracy of the trained Resnet-101 model in classifying ALL in microscopic images when the algorithm hyperparameter combinations in Table 4 were used in three independent experimental runs

Experiments 1–9	Dataset	Runs of experiment			Average accuracy	SD	η value
Experiments 1–9	Dataset	1	2	3	Average accuracy	SD	η value
1	Training set	0.9777	0.9796	0.9792	0.9788	0.0010	33.4870
1	Preliminary test set	0.8045	0.8066	0.7927	0.8013	0.0075	14.0346
2	Training set	0.985	0.9864	0.9872	0.9862	0.0011	37.2024
2	Preliminary test set	0.7916	0.7943	0.805	0.7970	0.0071	13.8487
3	Training set	0.9211	0.9218	0.9216	0.9215	0.0004	22.1026
3	Preliminary test set	0.7477	0.7483	0.7483	0.7481	0.0003	11.9754
4	Training set	0.7892	0.7888	0.7893	0.7891	0.0003	13.5185
4	Preliminary test set	0.6508	0.6508	0.6508	0.6508	0.0000	9.1385
5	Training set	0.9533	0.9538	0.9535	0.9535	0.0003	26.6572
5	Preliminary test set	0.7783	0.7809	0.7788	0.7793	0.0014	13.1253
6	Training set	0.864	0.8639	0.8647	0.8642	0.0004	17.3420
6	Preliminary test set	0.6909	0.6904	0.6888	0.6900	0.0011	10.1737
7	Training set	0.985	0.9877	0.985	0.9859	0.0016	37.0156
7	Preliminary test set	0.8056	0.8013	0.7965	0.8011	0.0046	14.0288
8	Training set	0.9056	0.9057	0.9064	0.9059	0.0004	20.5282
8	Preliminary test set	0.7268	0.7327	0.7338	0.7311	0.0038	11.4082
9	Training set	0.9796	0.9831	0.9831	0.9819	0.0020	34.8624
9	Preliminary test set	0.7954	0.7868	0.7563	0.7795	0.0205	13.1318

ISSN: 1471-2105