Performance enhancement of data mining algorithms for chronic diseases and medline documents

Abstract

Medical Data mining is the process of extracting hidden patterns from large medical data This thesis aims to develop a disease prediction model with enhanced performance and MEDLINE document classification model to improve the retrieval of the disease related information Different models are applied to the prediction of clinical data To design a disease prediction model an Amalgam kNN approach is proposed This classification model is proposed for chronic communicable as well as non communicable diseases datasets The datasets are collected from the University of California Irvine UCI Machine Learning repository GitHub and Kaggle data science repository In the Amalgam kNN model high dimensional data are handled by Correlation based Feature Selection CFS attribute reduction method and the kMeans algorithm identifies natural groupings among the observations To build the classifier kNN classification algorithm is used The proposed model produces one of the best Classification Accuracy of 97 4 for the PIDD data set and 99 7 for High Dimensional Diabetic readmission dataset and it is compared with simple kNN and other leading disease prediction models The proposed model is validated for all the chronic disease datasets and the results are compared with the leading clustering and classification algorithms It achieves better performance than Decision Tree J48 Support Vector Machine SVM Neural Network NN and Naive Bayes NB classification algorithms newline