Affiliations: [a]
Department of IT&CA, Sri Krishnan Adithya College of Arts and Science College, Coimbatore, Tamilnadu, India
| [b]
PG and Research Department of Information Technology, Government Arts College, Coimbatore, Tamilnadu, India
Correspondence:
[*]
Corresponding author. M. Pushpa, Department of IT&CA, Sri Krishnan Adithya College of Arts and Science College, Coimbatore, Tamilnadu, India. E-mail: pushpamphd11@gmail.com.
Abstract: The requisite of detecting Autism in the initial stage proposed dataset is exceptionally high in the recent era since it affects children with severe impacts on social and communication developments by damaging the neural system in a broader range. Thus, it is highly essential to identify this Autism in the primary stage. So many methods are employed in autism detection but fail to produce accurate results. Therefore, the present study uses the data mining technique in the process of autism detection, which provides multiple beneficial impacts with high accuracy as it identifies the essential genes and gene sequences in a gene expression microarray dataset. For optimally selecting the genes, the Artificial Bee Colony (ABC) Algorithm is utilized in this study. In contrast, the feature selection process is carried out by five different algorithms: tabu search, correlation, information gain ratio, simulated annealing, and chi-square. The proposed work utilizes a hybrid Extreme Learning Machine (ELM) algorithm based Adaptive Neuro-Fuzzy Inference System (ANFIS) in the classification process, significantly assisting in attaining high-accuracy results. The entire work is validated through Java. The obtained outcomes have specified that the introduced approach provides efficient results with an optimal precision value of 89%, an accuracy of 93%, and a recall value of 87%.
Keywords: Autism, data mining, gene expression, gene selection, hybrid classifier
