Implementation of Stacking Techniques to Improve the Performance of Children^s Autism Prediction Models Yulianti (*), Aries Saifudin, Irpan Kusyadi, Teti Desyani, Sri Mulyati, Endar Nirmala
Universitas Pamulang
Jl. Raya Puspitek No. 46 Buaran, Serpong, Tangerang Selatan, Banten, Indonesia, 15417
*yulianti[at]unpam.ac.id
Abstract
The problem of Autism Spectrum Disorder (ASD) is currently experiencing a rapid increase in all ages of the human population. Early detection and treatment of this neurological disease are very helpful in maintaining mental and physical health. Detection of ASD is a challenge because there are several other mental disorders whose symptoms are so similar to those of ASD that it becomes a difficult task. The speed and efficiency of the diagnosis of human health problems are very important. In Autism, the big challenge faced in many health care conditions is the timing of diagnosis. It can take up to 6 months to diagnose a child with autism with certainty due to the long process, and a child has to see many different specialists diagnose autism, from a developmental paediatrician, neurologist, psychiatrist or psychologist. In today^s traditional way, the time it takes to complete a diagnosis of autism is relatively long. With the widespread application of machine learning-based models in the prediction of various human diseases, early detection of ASD based on various health and physiological parameters has become possible. Many classification algorithms can apply to predict the presence of ASD. Each classifier is different in how it collects data, data filter, extract features, and uses this process to enter a model for study. In this study, the effectiveness of several machine learning algorithms to evaluate the predictive effectiveness of ASD. This research proposes the application of the stacking technique to predict the presence of ASD. Stacking is an efficient ensemble method, in which predictions are generated by combining several different machine learning algorithms. This research has shown that ASD prediction applying a stacking technique can provide better performance than applying single classifier.
