Distance Based Outlier Detection (DOD) with Gradient Tree Boosting Classifier Algorithm for Heart Attack Prediction

Abstract

This research discusses the evaluation and development of OD and classification model depends on recognizing the outliers in clinical care and classifying it. Outliers are the activities which are uncommon and might denote errors in the patient- management. The benefits of this methods are (i) to model a detection system, it does not need expert input (ii) empirically, the relative medical outliers are extracted by employing a huge set of history of patient cases and updated continuously to denote common practice patterns and iii) coverage of alert might be deep and extensive. For positive and extensive impact over medical care, this novel method comprises major potential. In this research, a novel technique named DOD is introduced for obtaining the patient's care only for Heart Attack, which has all patient-management performance that is relied on the state of patient. It leads to analyze the patient-management operation for given data that is more abnormal and similar to predefined patients. Once the difference has been noted, it is employed in classifying the patient details that tends to find Heart Attack. Here, GBTC method is applied for classification process. In order to guarantee the accuracy of projected DOD-GBTC method, a set of  2 standard dataset such as Heart Attack Statlog as well as Cleveland dataset on the basis of various computing metrics.