A hybrid Kernel probability based PSO Feature Selection model for breast cancer disease prediction

Authors

  • Anusha Derangula, Prof. Srinivasa Reddy Edara

Abstract

As the size of the biomedical databases increases day-by-day, finding an essential feature
set for classification problems is complicated due to large data size and sparsity problems.
Microarray feature ranking and classification are significant challenges to scientific and medical
researchers due to its high dimensional feature space and limited samples. Feature transformation,
feature ranking and data classification are the essential components to improve the microarray cancer
prediction on high dimensional datasets. It affects the physical, mental, social lifestyles of the people.
It is possible to treat cancer in the early stages. The importance of cancer cell classification into
benign and malignant has led to many research areas in the medical field. Medical practitioners were
adopting machine learning techniques to detect, classify, and predict the malignant tumour
effectively. The machine learning algorithms yield better results in the diagnosis of malignant tissue.
The learning algorithm performs well with optimal features. In this work, a hybrid kernel PSO based
feature selection approach is designed and implemented on the breast cancer datasets. The proposed
kernel based PSO model computes feature scores for classification problem. The subset of features is
used to predict the disease using different classification models such as Gradient Boosting, Extra
Trees, Naive Bayes, Random Forest and Support Vector Machine. The optimized features gave the
highest accuracy of 98% when implemented with Naive Bayes, Radom Forest and Support Vector
Machine classifiers.

Published

2020-12-01

Issue

Section

Articles