Hybrid Feature Selection and Heterogeneous Clustering Ensemble Framework for Detection of Circulating Tumor Cells

Breast Cancer (BC) is a major concern in huge part of women in western countries. Circulating Tumor Cells (CTC) especially for blood testing based on the genomic research investigation may be a capable way to solve BC detection, but it is restricted due to high dimensional gene features and missing data. In order to overcome these problems, the proposed work has been developed by using Decimal Scaling, Mix-Max and Z-Score Normalization schemas is applied for finding the missing values for gene samples. For preprocessing gene samples missing attributes data are imputated, and then the proposed method solves the genes selection problem by Fuzzy Online sequential Ant colony Kernel Extreme Learning Machine (FOA-KELM) schema. The FOA-KELM the mean values are computed for each gene feature via the use of ELM objective function to select the most important gene features. The Heterogeneous Clustering Ensemble Framework (HCEF) similarity measurement results are fused based on Weighted Quality (WQ), which in turn used to improve classification results.