Abstract
The neurological problem autism spectrum disorder, affecting many people globally, is multifarious in description; consequently, this makes quick detection of autism spectrum disorder (ASD) critical in the process of expert intervention. This work attempts to compare the use of multiple classifiers, logistic regression, k-nearest neighbor, and ensemble techniques such as random forest, CatBoost, and XGBoost. In the project, we have used Power BI visualizations to build a live interactive dashboard that will allow us to better understand the features present in the pre-acquired dataset and hence, help us with additional analysis and accuracy. Finally, the key findings of the project show that logistic regression did realize the test accuracy of 99.1 percent, while CatBoost, as well as random forest, got even better on a balanced dataset with the test accuracy of 99.5 percent and 99.6 percent, respectively. The results demonstrate that ensemble models are appropriate in labelling if someone has autism spectrum disorder (ASD). By conjoining analytical tools and machine learning approaches, one can presume smoother workability of project work for healthcare professionals, hence making the process comparatively less work intensive. Future work on the project may be to include advanced methodologies and to get somewhere close to developing an optimal model for clinical setups. This research work points out how amalgamating machine learning and data analytics could be beneficial for the advanced screening and diagnosis of ASD.