An improved ant colony optimization algorithm using the hellinger distance to predict high-dimensional and imbalanced bankruptcy data of shariah-compliant securities in Malaysia

Bankruptcy prediction is a critical area of study due to its significance in mitigating economic losses for stakeholders. However, the complexity and imbalance of bankruptcy datasets pose challenges to accurate prediction. This study addresses these challenges by utilizing attribute reduction techniques to manage high-dimensional data and identify optimal subsets for bankruptcy prediction. Additionally, since bankruptcy datasets are inherently imbalanced, with significantly fewer bankrupt companies compared to successful firms, conventional prediction algorithms often exhibit bias toward the dominant class. To overcome this limitation, this study proposes an enhanced Ant Colony Optimization (ACO) classification algorithm, named HDAntMiner, which incorporates Hellinger distance (HD) as a heuristic function. The HD mitigates bias toward the majority class, improving classification performance for minority instances.