ASD-Pipeline: An Ensemble Machine Learning Framework Integrating Feature Selection, Behavioural Clustering, and Class Rebalancing for Accurate Autism Spectrum Disorder Prediction

Approved

Relevance and Originality

The study introduces a compelling solution to the well-documented challenges in autism spectrum disorder prediction by merging multiple machine learning techniques into one structured pipeline. What sets it apart is the integration of adaptive feature selection and behavioral pattern clustering within the same framework, addressing known weaknesses in generalization and data imbalance. This approach significantly enhances relevance in the healthcare analytics domain and provides a meaningful contribution by filling a methodological gap in ASD detection models through ensemble learning and synthetic data augmentation.

Methodology

The research adopts a detailed five-phase design, starting with Min-Max scaling for normalization, which ensures consistency in data range. The use of FlexiFeat to blend filter-based, wrapper-based, and embedded feature selection techniques is a strong methodological highlight, reducing dimensionality and increasing relevance. Incorporating both K-Means and DBSCAN for behavioral grouping captures complex patterns, while the application of Cluster-SMOTE strategically balances the dataset. Finally, the ensemble classifier leverages diverse algorithms for improved robustness, making this methodology both exhaustive and technically sound.

Validity & Reliability

The study presents a convincing argument for the robustness of the ASD-Pipeline, reporting high performance across a variety of standard evaluation metrics. The use of soft voting among diverse classifiers ensures the system captures varying data behaviors, supporting model reliability. Furthermore, the inclusion of Matthews Correlation Coefficient alongside F1-score, precision, and specificity provides a balanced validation strategy. Nevertheless, the findings would be even more compelling if tested against independent datasets to strengthen claims of generalizability and confirm applicability in varied clinical settings.

Clarity and Structure

The progression from problem statement to technical implementation is clear, and the logical sequencing of the pipeline’s stages ensures reader comprehension. The research is well-articulated, and technical components are described with clarity and appropriate depth. Each technique used is justified within the broader context of autism prediction. While the narrative is precise, the article could benefit from the inclusion of visual process flows or schematic overviews to reinforce the modular architecture of the proposed system.

Result Analysis

The performance gains reported over previous methods are well-articulated and supported with a comprehensive set of metrics. The superiority of the ASD-Pipeline is effectively demonstrated through comparative accuracy, specificity, and F1-score. The results validate the strategic inclusion of behavioral clustering and synthetic oversampling as performance-enhancing components, and the conclusions are logically consistent with the data presented.