Automated Feature Selection in Microarray Data Analysis using Deep Learning

Cancer is the major reason of death around the world. However, timely identification and accurate prediction of cancer types play a pivotal role in safeguarding patient health. Microarray technology revolutionizes cancer detection by enabling the simultaneous examination of thousands of genes, simplifying the acquisition of extensive gene expression data. However, conventional cancer detection algorithms struggle with the vast amount of generated data. This study addresses these challenges by employing Principal Component Analysis (PCA) and Deep Learning, specifically Stacked Autoencoders, to reduce the dimensionality of large microarray datasets while preserving essential features. Utilizing these techniques, feature selection was conducted on datasets containing over 7000 features. The selected features were evaluated using standard regression and classification methods, with Logistic Regression emerging as the most effective, achieving an impressive 99.37% accuracy, followed by the Decision Tree classifier at 98.34%. Additionally, a Flask-based web application was developed to facilitate seamless CSV file upload for analysis, enhancing user accessibility and streamlining the data processing and analysis workflow. This user-friendly interface empowers researchers and practitioners to navigate through data complexities efficiently, fostering a more productive research environment.