Prediction Of Learning Disability Of The Children Using Adaptive Effective Feature Engineering Techniques

Abstract

Feature engineering is a critical step in the development of emergent machine learning models. Any process of selection and flexibility is included when using machine learning or mathematical modeling to develop a speculative model. One of the main objectives of predictive modeling is to find a reliable and accurate correlation between a set of available data and a given outcome. Machine learning classifiers are critical for detecting autism spectrum disorders early on. The purpose of this article is to raise awareness about the early detection of ASD in children who are affected. Autism spectrum disorder (ASD) refers to a set of conditions characterised by difficulties with social skills, repetitive activities, speech, and nonverbal communication. We proposed an adaptive CMR-ASD feature engineering model in this study that provides an effective technique for analyzing autism not only for doctors but also for psychologists and learning disability mentors. A hybrid adaptive CMR-ASD model combines various feature selection strategies such as the CHI²test, MUTUAL_INFORMATION, and RFE with PCA to pick a subset of important features by taking into account both the score and ranking of individual features. With appropriate feature selection, this enhanced model is utilized to predict autism in its early stages.  A real-time dataset, as well as four different datasets related to autism spectrum disorders, were used in the research. The results showed that the suggested strategy is capable of selecting highly disparate features, and the Matthews correlation coefficient (MCC) is a more reliable statistical rate that generates higher scores than the Cohen's kappa and accuracy scores.