Aggregate linear discriminate analyzed feature extraction and ensemble of bootstrap with knn classifier for malicious tumour detection

Abstract

Tumour detection medical applications utilize classification techniques to categorize malicious and non-malicious tumour features to provide an efficient medical diagnosis of the human individual under investigation. One way to enable efficient classification, Feature extraction methods are used to eliminate the redundant features and obtain the most relevant features. However, the challenges concerning the dimension and quantum of tumour dataset persist. Toward this goal, this paper aims to maximize the malicious tumour classification accuracy using two reliable ensemble classifiers namely Bootstrap Aggregation and k-nearest neighbour. Tumour features extracted by Aggregate Linear Discriminate Analysis (LDA) and the feature distance is calculated with iterative scattering matrix algorithm. The extracted features are further refined by aggregation to select most effective feature values. After this, an ensemble classifier technique is employed to construct malicious and non-malicious tumour classes. The tumour classification based on an ensemble of bagging and k-nearest neighbour. Simulation is carried out on Tumour Repository data set to show that proposed ensemble classifiers have considerably better tumour detection accuracy than existing conventional techniques. Numerical performance evaluations show that 8% improvement by proposed method in tumour classification accuracy for malicious tumour detection in human individuals.