A novel clinical decision support model for risk factor analysis in medical data

Abstract

The domain of medical informatics is concerned with the knowledge of information and computer science, engineering and technology towards health studies, medicine and its practices. The intersection among medical science and information technology provides the solution towards data-driven decision making. The future community is in need of integrated access to clinical information to formulate computer-based data management and decision-making process. The existing system mostly uses repository data access with evaluation from various classification techniques. Also, the algorithmic model has been evaluated only in terms of accuracy which doesn t support customization of all the datasets. The major consequence that has correlation and dependency among the attributes has not been determined over real world dataset to explore the undermined patterns and relationship in medical data. Feature selection also, in turn, uses up by some of the authors but the relevance, and investigation of selected attributes along with its contribution to the disease has not been measured and signified. In the context of the thesis, we are looking at feature selection using swarm intelligence techniques with data classification along with statistical inference for data validation, relevance, and feature set correlation. In addition, several researchers have worked on feature selection along with data classification with an explicit motto of improvement in accuracy. Since accuracy alone doesn t provide a valid solution to medical practitioners in evaluating and validating the developed model. The model should explicitly reveal the major set of risk factors, its relevance, correlation and feature subset evaluation with statistical data analysis. If the association and the relationship among the risk factors are noticed with the developed model it simplifies the process of revealing the relational aspects among medical expert The direction of interest to this thesis work lies in the utilization of swarm intelligence techniques upon multiple exhaustive searches followed by classification for the data collected from a regional hospital. The relationship, validation, and association among the selected subsets have been subjected to a thorough statistical data analysis. newline