Hybrid Multifilter Ensemble Based Feature Selection Model from Microarray Cancer Datasets Using GWO with Deep Learning
Bibhuprasad Sahu, Sujata Dash
Abstract
Clinical management plays an important role in cancer prognosis. The advancement of biological translational research and sophisticated statistical analysis with machine learning techniques are the driving factors behind enhancing cancer prognostic predictions. This study presents a hybrid multifilter-ensemble machine-learning model using Grey Wolf Optimizer (GWO) with Recurrent Neural Network and long-short term memory (LSTM) classifier. This model uses an ensemble feature selection with RNN and LSTM to classify microarray cancer datasets. Five different feature selection (FS) filter models such as information gain (IG), gain ratio (GR), chi-square, and correlation-based feature selection (CFS) used to identify the featured features from the original dataset. Finally, the ensemble multifilter selected features are classified using Recurrent Neural Network (RNN) and LSTM. The experiments are performed on leukemia, lung, and Small Blue Round Cell Tumours (SBRCT) datasets. The efficacy of the GWO-RNN and GWO-LSTM is performed using various performance metrics like precision, recall, accuracy, and F1_Score. The performance of the MF-GWO-RNN outperforms with an accuracy of 97.11%, 95.92%, 92.81%, and in the case of MF-GWO-LSTM outperforms with an accuracy of 97.17%, 98.56%, 96.38% with Leukaemia, Lung, SRBCT datasets respectively, which is better than MF-GWO-SVM, MF-GWO-KNN.