A hybrid approach using Naïve Bayes and Genetic Algorithm for childhood obesity prediction

Naïve Bayes is a data mining technique that has been used by many researchers for predictions in various domains. This paper presents a framework of a hybrid approach using Naïve Bayes for prediction and Genetic Algorithm for parameter optimization. This framework is a solution applied to the childhood obesity prediction problem that has a small ratio of negative samples compared to the positive samples. The Naïve Bayes has shown a weakness in prediction involving a zero value parameter. Therefore, in this paper we propose a solution for this weakness which is using Genetic Algorithm optimization. The study begins with a literature review of the childhood obesity problem and suitable data mining techniques for childhood obesity prediction. As a result of the review, 19 parameters were selected and the Naïve Bayes technique was implemented for childhood obesity prediction. The initial experiment to identify the usability of the proposed approach has indicated a 75% improvement in accuracy.