Ankylosing spondylitis prediction using fuzzy K-nearest neighbor classifier assisted by modified JAYA optimizer

Area of Science:

• Computational intelligence and machine learning applied to medical diagnostics.
• Optimization algorithms inspired by biological swarm behavior.

Background:

• Diagnosing ankylosing spondylitis (AS) is complex, involving medical history, clinical symptoms, and radiological evidence.
• Current diagnostic methods can be time-consuming and prone to inaccuracies, leading to delayed or missed diagnoses.
• There is a need for supplementary diagnostic techniques to improve AS detection and prognosis.

Purpose of the Study:

• To introduce an enhanced optimization algorithm, SCJAYA, for improved performance.
• To develop and validate a binary SCJAYA-based feature selection classifier (bSCJAYA-FKNN) for AS diagnosis and prognosis.

Main Methods:

• Developed SCJAYA by integrating salp swarm foraging and cooperative predation into the JAYA algorithm.
• Evaluated SCJAYA against 18 other meta-heuristic algorithms on 30 benchmark functions.
• Proposed the bSCJAYA-FKNN classifier using binary SCJAYA for feature selection, validated on 11 public datasets and an AS-specific dataset.

Main Results:

• SCJAYA demonstrated superior convergence speed and solution precision compared to conventional and state-of-the-art algorithms.
• The bSCJAYA-FKNN model achieved high performance metrics: 99.23% accuracy, 99.52% specificity, 0.9906 MCC, and 99.41% F-measure.
• The model achieved a computational time of 7.2800 seconds on the AS dataset.

Conclusions:

• The SCJAYA algorithm offers enhanced optimization capabilities.
• The bSCJAYA-FKNN classifier shows significant promise for accurate and efficient diagnosis and prognosis of ankylosing spondylitis.
• This computational approach can potentially reduce the clinical burden associated with AS diagnosis.