Genes expression Dimensions reduction for Cancer classification based on optimized deep neural network with Giza pyramid construction

Abstract High-dimensional data, such as microarray data, are commonly utilized to diagnose diseases. In this type of data, each array corresponds to a gene in the chromosomal makeup. As certain diseases are caused by gene mutations, identifying these genes is crucial for accurate disease classification. Feature selection (FS) is a key approach in pattern recognition and bioinformatics to reduce the number of dimensions in a dataset. However, selecting a subset of features that maintain the original data's characteristics without sacrificing classification accuracy is a challenging task, as it is an NP-hard problem. Meta-heuristic optimization methods have shown promising results in addressing this issue. This paper proposes a feature selection approach that employs the Giza Pyramids algorithm with a deep learning kernel to identify informative genes for cancer patient classification. The method is evaluated using five well-known microarray datasets in the field of cancer diagnosis. The experimental results demonstrate that the proposed method outperforms other classification methods in terms of various evaluation criteria. Specifically, the Giza Pyramids algorithm with a deep learning core successfully selects useful genes for cancer patient classification, leading to improved classification accuracy.