Implementation of ensemble machine learning algorithms on exome datasets for predicting early diagnosis of cancers

Table 2 Performance metrics for KNN, SVM, MLP and ensemble classifiers for five cancer types

Classifier	Cancer class	Precision	Recall	F1-score
(a) Performance metrics for KNN
KNN	High-grade serous ovarian cancer	0.56	0.61	0.59
	Human diffuse type gastric cancer	0.88	0.71	0.79
	Intrahepatic cholangiocarcinoma	0.79	0.85	0.82
	Non BRCA1/BRCA2 familial breast cancer	0.82	0.96	0.89
	Pancreatic adenocarcinoma	0.60	0.62	0.61
Weighted accuracy		0.77
(b) Performance metrics for SVM
SVM	High-grade serous ovarian cancer	0.66	0.58	0.62
	Human diffuse type gastric cancer	0.83	0.66	0.73
	Intrahepatic cholangiocarcinoma	0.85	0.86	0.86
	Non BRCA1/BRCA2 familial breast cancer	0.84	0.99	0.91
	Pancreatic adenocarcinoma	0.62	0.71	0.66
Weighted accuracy		0.76
(c) Performance metrics for neural networks
Neural networks	High-grade serous ovarian cancer	0.75	0.74	0.74
	Human diffuse type gastric cancer	0.83	0.78	0.80
	Intrahepatic cholangiocarcinoma	0.85	0.89	0.87
	Non BRCA1/BRCA2 familial breast cancer	0.89	0.92	0.91
	Pancreatic adenocarcinoma	0.78	0.78	0.78
Weighted accuracy		0.82
(d) Performance metrics for ensemble model
Ensemble model	High-grade serous ovarian cancer	0.76	0.78	0.77
	Human diffuse type gastric cancer	0.82	0.77	0.79
	Intrahepatic cholangiocarcinoma	0.84	0.91	0.87
	Non BRCA1/BRCA2 familial breast cancer	0.89	0.93	0.91
	Pancreatic adenocarcinoma	0.83	0.77	0.80
Weighted accuracy		0.82

ISSN: 1471-2105