LAUSR

Microarray data analysis is a most promising and difficult process due to the complex nature of data. It includes higher dimensionality, several unbalanced classes, a smaller sample size, the presence of noise, and a higher variation of feature values. This has resulted in a decrease in classification accuracy as well as an overfitting problem. This work proposed an efficient and hybrid deep learning technique for molecular cancer classification using expression data to solve these limitations. The different steps in the proposed work are preprocessing, clustering, extraction, selection, and classification. The input data is preprocessed using a scalable range adaptive bilateral filter. Then clustering is done with the help of an improved binomial clustering approach. After that, the data is extracted with the help of the multifractal Brownian motion method. Then the important features are selected with the help of an improved cuckoo search optimization algorithm. Finally, the data classification is performed using a wavelet‐based deep convolutional neural network. This work is validated with the help of five publically available datasets using the PYTHON platform. The different performance measures considered here are accuracy, precision, recall, and F‐measure. The classification accuracy obtained is 98.36%, 98.12%, 98.55%, 97.70%, and 95.30% for ovarian, breast, colon, leukemia, and prostate cancer datasets. The overall result showed that the suggested technique is better than the existing methods.