A Survey on Cancer Classification with Deep Learning from Microarray Gene Expression Data
Bioinformatics is a field that deals with the analysis of biological data, under this selection of genes based on extreme-quantity biological data (microarrays) is a critical challenge. Gene expression records typically contain a small number of samples many redundant and irrelevant genes. Inappropriate and redundant genes degrade the learning models’ performance (classifiers). By reducing the number of dimensions and improving the prediction accuracy of classifiers, choosing high discriminatory subsets of genes derived from microarray data can help minimize computational expenses. The goal of gene selection is to increase the quantity of genes into population and identify a small sample of genes with the highest discriminating power to increase prediction performance while retaining strength. This knowledge aids physicians in clinical preparation in making quick and accurate diagnoses and treatments. Gene selection is an effective data pre-processing strategy in cancer classification. Microarray data analysis yields useful results that aid in the resolution of gene expression outline issues. Among the highly significant Cancer classification is a usage of microarray data analysis. Microarray's dataset is high-dimensional, meaning it comprises thousands of genes (features) and has minimal data sparseness. Several tens of samples are generally used. The intricacy of gene expression data is high; genes relate to each other directly or indirectly. Creating a new ensemble feature selection algorithm to improve feature selection accuracy and dependability. As deep learning algorithms work well when we have smaller number of samples and more features which is appropriate in our case. Hence deep learning is considered rather than machine learning in order to get more accurate results.
Keywords - Gene Selection, Micro-Array, Gene Expression, Bio- Inspired, Feature Selection, Cancer Classification.