Feature Selection of Microarray Data For Cancer Detection Using Simulated Kalman Filter (SKF) With Mutation
Over the last decades, gene expression has received increasing attention by scientists due to the development and advancements in deoxyribonucleic acid (DNA) microarray technology. DNA microarray technology allows us to measure the expression levels of a large number of genes simultaneously. Microar...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2019
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Over the last decades, gene expression has received increasing attention by scientists due to the development and advancements in deoxyribonucleic acid (DNA) microarray technology. DNA microarray technology allows us to measure the expression levels of a large number of genes simultaneously. Microarrays have been proven useful when it comes to understanding the genetics of a disease. It has been used to assess many different types of cancer that is not limited only to single cancer types but also between multiple human malignancies. Nevertheless, DNA microarray data usually contain thousands of genes and most of them are proving to be uninformative and redundant. Numerous studies have shown that most genes measured in DNA microarray experiments are not contributing to the accuracy of classification. Thus, there is a need of feature selection which allow high predictive accuracy and better diagnostic performance. There are various feature selection techniques available to enhance the classification of DNA microarray. This study highlights the pioneer application of estimation-based metaheuristics for feature selection of microarray data. Different from SI, EA and other inspired algorithms, this estimation-based algorithms solves optimisation problems by estimating an optimal solution. Simulated Kalman Filter (SKF) is applied for feature selection of gene expression data. A new variant of SKF which is Simulated Kalman Filter with Mutation (SKF-MUT) is introduced to enhance the performance of original SKF in the selection of microarray features. Benchmark datasets were used which considered both binary and multiclass datasets. Artificial Neural Network (ANN) is used to study the performance of the feature selector proposed. The accuracy of the prediction is taken as the performance measurement by considering the confusion matrix. Experiments were conducted by using MATLAB 2018a with Neural Network Toolbox. Based on the results, SKF and SKF-MUT managed to effectively select the number of features needed for higher accuracy of classification randomly. The proposed algorithms are compared with existing works that used the same benchmark datasets. Statistical analysis using Friedman test showed that SKF-MUT is ranked first and on par with the other benchmark existing studies. |
|---|