Classification of Preterm Birth with Optimized Rules by Mutating the Cognitive Knowledge of PSO

Background/Objectives: Classification of preterm birth is a major challenge in the absence of effective tools and lack of domain knowledge with minimized set of rules. Objective of this paper is to classify the preterm birth with optimized rules. Methods/Statistical Analysis: This paper, proposes mutated Particle Swarm Optimization in order to achieve better result. Cognitive component was obtained by applying the mutation technique followed in the evolutionary algorithm. Mutation phase helped in converging to the optimized solution much faster. Findings: Preterm dataset of 1052 records with 5 attributes was applied to the proposed algorithm. Mutation was applied based on the Poisson, Gaussian, Uniform and Exponential distributions. The result shows that Poisson mutation among other distributions applied on the personal best has reduced the number of rules needed for classifying preterm birth datasets for both training and testing. It was observed that the optimized 13 rules out of 28 for training dataset and 9 rules out of 28 rules wereonly needed for classification. Application/Improvements: The proposed algorithm was suitable for preterm birth data set. This can be applied on the dataset with medium initial population. It was observed thatthe efficiency of the algorithm varied depends on the type of data. Therefore a generic method can be developed.