NAÏVE BAYESIAN CLASSIFICATION OFGESTATIONAL DIABETES MELLITUS USING MODIFIED PARTICLE SWARM OPTIMIZATION

Abstract

Gestational Diabetes Mellitus - GDM is a public health problem. With changes in eating habits, increased purchasing power, and climate change, among others, the number of women with gestational diabetes complicated by pregnancy is increasing. GDM generates problems for the mother and for the baby. Therefore, early diagnosis is important for adequate medical follow-up and treatment in a timely manner. Feature selection is an important process in data mining to extract features when the number of features is large.  Asthere are 2ⁿfeature subsets for n number of featuresin feature selectionevery feature has two possibilities. The first one is that particular feature would be selected for classification and the other is it would not be selected for classification. So,finding a relevant feature subset in appropriate time is a NP-Hard problem. To avoid this problem, the approximation algorithm is used that gives the near optimal solution. There are four types including filter, wrapper, embedded and hybrid techniques. Many of the swarm intelligent algorithms that simulate the social behaviour of living beings are used as feature selection algorithms.   Naïve Bays approach is so sensitive to change their parameters. A modified Particle swarm optimization is used for feature selection (PSO) and the Naïve Bays Classifier (NBC) to classify the dataset with the selected features.  The proposed approach named Modified PSO-NBCapproach is implemented with PIMA Indians data set collected from UCI repositories of machine learning database. The performance of the proposed approach is compared with the existing methods. The experimental study shows that theModified PSO-NBC minimizes the computation cost and  time, maximizes the ROC and classification accuracy than several other existing methods.