Jaya Optimized Extreme Learning Machine for Breast Cancer Data Classification
| Autor: | Santos Kumar Baliarsingh, Swati Vipsita, Chinmayee Dora |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Breast biopsy
Artificial neural network medicine.diagnostic_test Wilcoxon signed-rank test Computer science business.industry Data classification Overfitting Machine learning computer.software_genre medicine.disease Support vector machine Breast cancer medicine Artificial intelligence business computer Extreme learning machine |
| Zdroj: | Smart Innovation, Systems and Technologies ISBN: 9789811562013 |
| Popis: | As the latest World Health Organization (WHO) statistics show, approximately 15% of all cancer deaths among women, i.e., 627,000 in number in the year 2018 are from breast cancer in the USA. To make a definitive cancer diagnosis, a biopsy is a medical practice for maximum types of cancers, as it provides the most accurate analysis of tissue. To avoid the huge number of undesirable breast biopsies, new diagnosis methods based on microarray data analysis have been proposed in the last decades. The microarray analysis can help the physicians to decide whether to carry out a breast biopsy or not. Gene expression profiling based on microarray data has been evolved as an effective procedure for the classification of cancer along with its diagnosis and treatment. Machine learning techniques like artificial neural network have shown significant potential in cancer classification and clinical diagnosis. In this paper, a classification method based on Jaya optimized extreme learning machine (JELM) has been applied on breast cancer microarray data after relevant genes are selected employing Wilcoxon rank sum test. Jaya is used to pre-train the ELM by selecting the optimal input weights and hidden biases. ELM rectifies the difficulties raised by iterative learning techniques such as local minima, improper learning rate, and overfitting. Finally, a comparative result analysis is presented on the achieved classification accuracy by four well-known classifiers available in the literature, namely SVM, kNN, NB, and C4.5. Accuracy is used as a performance metric to analyze the efficiency of the classifiers. From the obtained results, it is observed that JELM classifier achieves better classification accuracy as compared to other schemes. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|