Categorization of self-care problem for children with disabilities using partial swarm optimization approach

Abstract

Self-care or personal care denotes those actions or doing that a person undertakes in supporting personal health, limiting personal illness, preventing personal disease and reinstating their own health. Self-caring is big challenge in exceptional/disabled children. With recent advancement in artificial intelligence in last few years, machine learning can be used for classification of self-care problem in children with different age groups. The paper proposed an enhanced expert system based on machine learning for diagnose and classification of self-care issues in children with physical and mental disorder. Partitioned Multifilter with Partial Swarm Optimization (PM-PSO) is used for attribute/feature selection and the outcomes are analogize with Principal Component Analysis (PCA). The preferred features/attributes are tested, trained and validated on following classifiers:-Naïve Bayes, Multilayer Perception (MLP), C-4.5 and Random Tree. tenfolded cross validation is used for validation, testing and training. PCA selects 32 attributes and shows truly categorized instances i.e. accuracy as: (1) 80% for Naïve Bayes; (2) 68.57% for MLP; (3) 68.57% for C 4.5 and; (4) 64.28% for Random Tree. The classifiers show a significant improvement in performance with PM-PSO feature selector. 50 attributes were selected with PM-PSO. It shows truly categorized instances/accuracy as: (1) 81% for Naïve Bayes; (2) 80% for MLP; (3) 80% for C 4.5 and; (4) 78.57% for Random tree.