Ensemble Classifier with Hybrid Feature Transformation for High Dimensional Data in Healthcare

In recent years, the growth of high-dimensional data applications has posed a significant challenge to data analytics, particularly high-dimensional data classification. To solve this problem, efficient feature selection and extraction approaches are paramount to the success of classification analyses. The dimensionality problem of high-dimensional data, which comprises a large number of variables and complicated data matrices, is a big barrier in the health care research domains. We propose an Ensemble Classifier with Hybrid Feature Transformation for handling extremely complex dimensional data. This is proven to be efficient for selecting and extracting features from a very large dataset in a classification task. The proposed approach has been applied to the classification of an unbalanced dataset of gene expression RNA-Seq data in five types of cancer. In the hybrid feature transformation approach, the original datasets are transformed in two stages. The first phase selects the highest score features from the original dataset, which are then sent to the second phase of feature transformation. As part of the second phase, features are transformed into a new set of features. Finally, a dataset with six features is applied to the ensemble classifier which achieves 99.96% accuracy by averaging the probability systems of the groups of classifiers. The findings clearly show that the proposed approach outperforms the existing methods for high-dimensional data classification.