Comparative Analysis of PSO and ACO Based Feature Selection Techniques for Medical Data Preservation

Sensitive medical dataset consist of large number of disease attributes or features, not all these features are used for diagnosis. In order to preserve the medical dataset it is not essential to perturb all the features before it is shared for mining purpose. To reduce the computational cost and to increase the efficiency, in this work tried to use Ant Colony Optimization (ACO) for feature subset selection which is used to reduce the dimension and also compared with feature subset selection using Particle Swarm Optimization (PSO) which is also used to reduce the dimension. Both the techniques are explored to reduce the dimension before applying preservation technique. By using randomization method a known distribution is added to the reduced sensitive data before the data is sent to the miner. The approach is analyzed using standard UCI medical datasets. The result is analyzed based on classification accuracy using machine learning algorithms (Naïve Bayes, Decision Tree) build on the randomized dataset. The experimental results show that the accuracy is maintained in the reduced perturbed datasets. The results also show that ACO search based feature selection has more accuracy than PSO search based selection.