An Improved Feature Selection (IFS) Algorithm for Detecting Autistic Children Learning Skills

M. S Mythili¹ and A. R Mohamed Shanavas²

¹Department of Computer Applications, Bishop Heber College, Tiruchirappalli - 620 017, TamilNadu, India. ²Department of Computer Science, Jamal Mohamed College, Tiruchirappalli - 620 020, TamilNadu, India.

ABSTRACT: Children with autism spectrum disorder follow a special process pattern from other children and develop at a special rate. Educating children with autism is expounded to intensive endeavor, involving a team of professionals of assorted special instruction and therapies to subsume children behavioral and activity, biological process, social and academic needs. Students with autism usually need explicit teaching across a range of settings to generalize skills. The knowledge and information gained from analysis helps the parents and school teachers to convey decent learning surroundings for the autism children. Feature Selection is a very important topic in data mining, particularly for high dimensional datasets. Feature Selection is a method usually employed in machine learning, whereby subsets of the options accessible from the data are described for the application of a learning algorithm. The Main purpose of this paper is to propose the Improved Feature Selection algorithm. The Improved algorithm is developed by combining the filters and wrappers. The Correlation Based Feature Selection (CFS) with the best first search act as a filter for removing impertinent options. Wrapper Subset Evaluator with the best first search is employed as a wrapper and it absolutely reduces the redundant options. It is accustomed to improve the accuracy of the classification for the autism children by analyzing the four totally different classifiers such as SVM, J48, Multilayer Perceptron and IB1are used. The overall purpose of this paper work is to propose the foremost effective algorithmic program for autism children with sensible potency and accuracy.

KEYWORDS: Data Mining; Classification; Filter Approach; Feature Selection Algorithms

This work is licensed under a Creative Commons Attribution 4.0 International License.