Classification of cancer microarray data using a two-step feature selection framework with moth-flame optimization and extreme learning machine

• Published in: Multimedia tools and applications 2024-02, Vol.83 (7), p.21319-21346
• Main Authors: Sucharita, Swati, Sahu, Barnali, Swarnkar, Tripti, Meher, Saroj K.
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Cancer; Classification; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Feature selection; Gene expression; Machine learning; Multimedia Information Systems; Optimization; Performance evaluation; Performance indices; Special Purpose and Application-Based Systems; Track 2: Medical Applications of Multimedia
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-16353-2

Analysis of microarray gene expression data for the detection/classification of cancer is one of the common approaches adopted worldwide. However, many genes (features) with correlated and irrelevant information in these data sets become the bottleneck for a classification model and significantly deteriorate its performance. A large number of features with fewer samples further make the classification task more cumbersome. Several feature selection methods (both filter and wrapper) have been proposed individually to address this issue, but choosing the best one among them is an open challenge. Our objective in the present study is to simplify the search for the best feature selection method without relying completely on individual methods and propose a two-step hybrid approach. In the first step, we use an ensemble of filter-based heterogeneous feature selection methods. These selected features then undergo the second step of wrapper-based selection. We propose to use the bio-inspired method called Moth-flame optimization (MFO) with an extreme learning machine (ELM) as its fitness function in this step. The motivation for using ELM is to leverage its learning strategy with one-pass processing of samples. Using this hybrid feature selection method, we proposed a classification model for Cancer Micraoarray data, where ELM is also considered as a classifier. The work demonstrates the superiority of the proposed model over other state-of-the-art methods in classifying cancer data from four different microarray gene expression datasets. Several measurement indexes are used for the performance evaluation of models.