An Analysis of Autism Disorder Factors using Different Classification Techniques

The healthcare environment is generally recognized as being information rich, with a knowledge poor. Despite the existence of a huge amount of data collected in medical records, however, the trends, patterns, and relationships between the crucial causes of Autism Disorder (AD) remain undetected. Knowledge of the risk factors associated with AD helps healthcare professionals to identify patients at high risk of having the AD. The aim of this research is to use data mining techniques, namely classification, to analyze and understand risk factors for the incident of autism in Saudi Arabia. This is achieved through building a classification model that predict the chances of having an autistic child for certain parents. This research contributes to emphasizing 13 risk factors that have been considered as potential for the occurrence of autism in previous studies. Two approaches of classification were used; incremental and non-incremental learning algorithms. The performance of different classification algorithms was compared in terms of classifier's accuracy, precision, and recall. IBK algorithm produces the better average accuracy for the classifier followed by RandomForest which is 80.6% and 78.3% respectively. Implementation steps and the main user interfaces of the proposed system are presented. Finally, some interesting rules were derived from J48 Decision Tree.