Computational Intelligence Data Analysis for Decision Support and Health Care Monitoring System

Abstract

Monitoring is a process of continuously gathering data and performing real-time analyses, monitoring can improve the estimation of the current state, identification of the critical situation, and assisted in decision support and planning. The past few years have an increase in the development of Ambient Intelligence health monitoring systems. An important aspect of investigation in such system is how the data is treated and analyzed. The survey of literature in this area presents that data mining analysis is lacking in Ambient Intelligence healthcare monitoring management. Though, there is good understanding of the importance healthcare systems by various authors, their focus was limited to a single aspect of the whole system and without integrated the analysis and decision support using machine learning and data mining methods. Hence, The goal of this paper is devoted to extensive investigation to construct a new novel ensemble health Care decision support for assisting an intelligent health monitoring system and also focus was to reduce the dimensionality of the attributes. Also the paper aims to discusses the findings of machine learning experiments and trend analysis on the simulation wearable sensors patients monitoring data. In the process of addressing the objectives of the paper indicated above, two major phases of experiments were conducted. In the first phase experiment, attempt has been made to investigate the experimental results of the performance of different classification techniques for classifying the data from different simulated wearable sensors used for monitoring different patients with different diseases. So as to construct the Base Classifiers Proposed used in the first experiment are: In the second phase experiment, we investigated various Meta classifiers. Finally new Novel Intelligent Ensemble method was constructed based of Meta classifier voting combining with three base classifiers J48, Random Forest and Random Tree algorithms. Different comparative analysis and evaluation were done using various evaluation methods like Error Metrics, ROC curves, Confusion Matrix, Sensitivity, Specificity and the Cost/Benefit methods. The results obtained show that the Novel Intelligent Ensemble method classifier is very efficient and can achieve high accuracy and, better outcomes that are significantly better compared with the outcomes of the all base classifiers proposed and all meta base classifiers.