POINT BISERIAL CORRELATIVE FEATURE PROJECTION BASED RELEVANCE VECTOR SOFTBOOST FOR DELIVERY MODE CLASSIFICATION

Abstract

Data mining is mainly used in the healthcare industry to predictpatient health conditions. It uses the data analytically to predict the most accurate delivery mode in advance for pregnant women. The existing methods performthe delivery prediction but it faces various changes while analyzing the risk factors. In order to perform accurate delivery prediction, an efficient technique called Point Biserial Correlative Feature Projection-based Relevance Vector SoftBoosting Classification (PBCFP-RVSBC) technique is introduced. The proposed PBCFP-RVSBC technique comprises the two major processes namely feature selection and classification. At first, the feature selection is carried out using Pointbiserial correlative projection pursuit to map the relevant features for predicting the deliverytypes with minimum time. After that, the SoftBoostensemble classification technique is applied for analyzing the risk factors associated with the cesarean by constructing several weak learners. The Jaccardkernelized relevance vector is used as a weak learner for analyzing the risk factors with training and testing data. Then the SoftBoost technique uses a totally corrective property to improve the classification accuracy and minimize the generalization error. Experimental evaluation is carried out with different factors such as accuracy, sensitivity, specificity and prediction time with respect to a number of patient data. The observed result shows that the presented PBCFP-RVSBC technique achieves higher accuracy, sensitivity, specificity and minimizes prediction time than the state-of-the-art methods