A novel framework for efficient medical data classification using hybridization of dimensionality reduction algorithms

Abstract

Data classification is effectively used in many fields such as Scientific experiments, Medical Industry, Credit approval, Weather Forecasting, Customer Segmentation, Marketing, Fraud Detection, Diagnosis, and Forecasting of various diseases in Biomedical Science. Among the diseases in the Medical domain, Cancer is one of the leading causes of death worldwide. Early detection of malignancy helps to reduce the mortality rate. Accurate prediction of Cancer disease can help in providing better treatment and minimised toxicity in the patients. Therefore, Cancer disease classification is proposed to address this challenge. In this work, two contributions are developed to implement the Medical data classification. In the first contribution, Hybrid Local Fisher Discriminant Analysis (HLFDA) based dimensionality reduction for Cancer disease prediction is developed. This approach consists of two main stages, HLFDA based dimensionality reduction and Type2fuzzy Neural Network (T2FNN) based data classification. The high dimensional data are significant obstacles for classification and these data also increase the computation complexity. To avoid this issue, HLFDA is utilized for dimension reduction and T2FNN is used for prediction. Further to enhance the classification accuracy, Feature Selection combined with Hybrid Support Vector Neural Network (HSVNN) based Medical data Classification is proposed in the second contribution. In this approach, relevant features are selected using the Adaptive Artificial Flora (AAF) Optimization Algorithm and Classification is performed using Hybrid Support Vector Neural Network (HSVNN) Classifier. For experimental analysis, three newlinedatasets such as Breast Cancer dataset, Cervical Cancer dataset, and Genetic expression Cancer dataset are utilized. Performance of both the contributions is analysed in terms of accuracy, sensitivity, and specificity. The proposed framework has achieved significant improvement in the Cancer disease prediction newline newline