Fig 1.
Gene selection using PGSA.
Fig 2.
Pseudo code of gene selection using pyramid IBGSA.
Table 1.
Characteristics and survival information for subgroups.
Table 2.
The overall accuracy of gene selection algorithms.
Fig 3.
Confusion matrix of optimization algorithms.
Fig 4.
Accuracy (%) of PGSA during thirty independent runs.
The Y and X axes imply the accuracy and iteration respectively. The bubble size is correlated with the number of genes; the bigger the bubble, the higher the number of genes. The best model was reached at the 18th run with 84.5% accuracy and 73 genes.
Table 3.
Five-fold cross-validated TPR, PPV, and F1-score of different algorithms.
The best result for each class has been bolded.
Fig 5.
Bottleneck subnetwork constructed based on PGSA selected genes in breast cancer.
Red and yellow colors indicate higher and lower bottleneck scores in the network, respectively.