Optimal Feature Selection using Particle Swarm Optimization with Random Forest Classifier for Lymph Diseases Prediction

Abstract

ML-based data classification approaches can be used as a decision making tool in various fields such as healthcare, disease prediction, etc. Presently, most of the data in medical domain comprise nature of high dimensionality Sometimes, FS (FS) methodologies is employed to improvise the classification results especially for high dimensionality issue by extracting the appropriate training instances to more number of determined features. This paper made an attempt to study the application of FS approaches on the classifier performance. For FS, genetic algorithm and particle swarm optimization (PSO) algorithm is used whereas random forest (RF) classifier is used for classification lymph diseases dataset. In the first stage, GA and PSO are used to reduce the feature subset and in the second stage, RF classifier is used. From the experimentation part, it is evident that the PSO based FS increases the classifier results compared to GA based FS. It is also studied that the FS process improvises the classifier results in a significant manner interms of diverse performance measures.