Hybrid Multiple Filter Embedded Political Optimizer for Feature Selection

• Published in: 2022 International Conference on Intelligent Controller and Computing for Smart Power (ICICCSP) pp. 1 - 6
• Main Authors: Sahu, Bibhuprasad, Panigrahi, Amrutanshu, Rout, Saroja Kumar, Pati, Abhilash
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.07.2022
• Subjects: Computational modeling; Feature extraction; Feature selection; Filter; Filtering algorithms; Metaheuristics; Microarray dataset; Multi filter; Political Optimizer; Power filters; Statistical analysis; Support vector machines; wrapper
• DOI: 10.1109/ICICCSP53532.2022.9862419

DNA microarray data analysis is notorious because it contains a large number of characteristics, asymmetrical class distribution, and it is restricted with less number of samples. We emphasize high-dimensional multi-class unbalanced issues in this work. The high dimensional, multi-class unbalanced problem has found it challenging for traditional classifiers to execute classification assignments effectively on both minority and majority classes. Numerous approaches have been made to handle either high dimensionality datasets or difficulties with class imbalance. Despite this, because of their complex relationships, few techniques have been presented to address the intersection of multi-class unbalanced and high-dimensional issues at the same time. Using multiple filter-based rankers (MF) and a hybrid political optimizer (PO), this study proposes unique hybrid algorithms for feature selection with the high-dimensional multi-class imbalanced issue. To evaluate the performance of the model, we have used four different classifiers named Support vector machine (SVM), Naive Bayes(NB), Decision Tree(DT), and K Nearest Neighbour(KNN). The experimental results show that the proposed methods are more effective than other well-known methods in the case of classification performance improvisation. Performance metrics accuracy indicates that the proposed methods are capable of searching the feature space and identifying very robust and discriminative features that best predict the minority class.